Article Text
Abstract
Background International evidence indicates relationships between pre-pregnancy body mass index (BMI) and breastfeeding behaviours. This study aims to assess associations between key points in the breastfeeding trajectory (initiation, early cessation and longevity) and pre-pregnancy BMI in a recent, nationally representative British cohort. It also aims to explore in the British context potential moderation by mothers’ ethnic group.
Methods The sample comprises 17 113 mothers from the UK Millennium Cohort Study who have information on pre-pregnancy BMI. Associations between pre-pregnancy BMI categories and breastfeeding initiation, early cessation and longevity are tested using logistic regression. Directed acyclic graphics identify appropriate minimal adjustment to block biasing pathways and classify total and direct effects.
Results After adjusting for confounders, there are large differences in breastfeeding early cessation and longevity by pre-pregnancy BMI group. Differences in propensity to initiation are negligible. Having begun breastfeeding, overweight and obese mothers are more likely to cease in the first week and less likely to continue past 4 months. Observed potential mediators within pregnancy and delivery provide little explanation for relationships. Evidence for moderation by ethnicity is scant.
Conclusions The causal mechanisms underlying relationships between pre-pregnancy overweight, obesity, and breastfeeding behaviours require further research. However, this study suggests pre-pregnancy BMI as one predictive measure for targeting support to women less likely to establish breastfeeding in the early days, and to continue beyond 4 months. The nature of support should carefully be considered and developed, with mind to both intended and potential unintended consequences of intervention given the need for additional investigation into the causes of associations.
- breastfeeding
- obesity
- perinatal
- longitudinal
Statistics from Altmetric.com
Introduction
Breastfeeding is dually advantageous for infant health and for maternal health,1 but despite its heavily promoted benefits, only a very small percentage of mothers in the UK meet the current WHO recommendation of exclusive breastfeeding for the first 6 months of life.2 The 2010 Infant Feeding Survey indicates that while a high percentage of mothers in the UK initiate breastfeeding (81%), fewer than one-fifth of babies are breast fed exclusively to 3 months (17%) and very few (1%) are breast fed exclusively to 6 months.3 Moderate increases in breastfeeding rates have the potential to save the National Health Service (NHS) £40 million per annum,4 but more research is needed to identify the modifiable factors associated with breastfeeding behaviours.2
There is accumulating evidence for the relationship between breastfeeding and body weight for both children and mothers. Exclusive breastfeeding is a viable means of reducing the risk of childhood overweight and obesity.5 Breastfeeding can also result in a quicker return postpartum to normal body mass index (BMI) in the mother.6 7 Yet, epidemiological studies provide consistent evidence of lower breastfeeding initiation and/or duration for mothers who were overweight and obese prior to pregnancy.8–12 Moreover, some suggest that the relationship between BMI and breastfeeding may be moderated by ethnicity,8 11 and there is emergent evidence that underweight may also be negatively correlated with breastfeeding.8
Despite the availability of rich contemporary sources of data in the UK, there is very little research on maternal BMI and breastfeeding behaviours specific to the UK context, with its distinctive ethnic composition and social/medical environment. Though Millennium Cohort Study (MCS) data have been used to provide evidence for maternal pre-pregnancy BMI influencing both children’s cognitive performance,13 and childhood BMI,14 neither it nor any of the other UK cohorts has, to the authors’ knowledge, been used to investigate pre-pregnancy maternal body composition and breastfeeding behaviours.
Given associations evidenced in the UK between breastfeeding and cognitive development,15 16 and breastfeeding and child BMI,17–19 it is possible that breastfeeding mediates the previously observed relationship between maternal pre-pregnancy BMI and child development. So exploring explicitly and quantifying the associations between BMI and breastfeeding in this UK sample will help inform the evidence base on what may be instrumental and modifiable in influencing breastfeeding and, potentially, in affecting children’s outcomes.
A number of possible confounders and mediators may interplay with the relationship between pre-pregnancy BMI and breastfeeding behaviours. Directed acyclic graphics (DAGs), graphical representations of conceptual models making explicit the proposed relationships between variables,20 can help to distinguish between these factors. A growing number of articles explain the benefits and logic of using DAGs,20–22 which can aid with the problem of overadjustment bias, and calculate the minimum amount of adjustment necessary to estimate the effect of the exposure on the outcome while minimising this bias.23
Confounders may relate separately to both unhealthy BMI and breastfeeding non-initiation/curtailed duration,20 and examples in the literature include age, ethnicity, parity, maternal smoking behaviour and maternal education.8 9 11 Mediators include those variables that can be caused by pre-pregnancy maternal BMI and potentially subsequently effect breastfeeding behaviour20: delivery method, birth weight, gestational age, gestational diabetes and gestational weight gain.8 9 11
Previous studies that have considered the relationship between maternal pre-pregnancy BMI and breastfeeding behaviour have described all of these variables, including potential mediators, as confounders, and treated them as such in their discussion of adjusted results.8 9 11 In contrast, the current study improves on previous research by separating mediators and confounders at the outset, and examining the total effect and direct effect of maternal pre-pregnancy BMI on breastfeeding behaviour.
The aim of this study is therefore to provide evidence on the relationship between maternal pre-pregnancy BMI and breastfeeding initiation, early cessation and longevity, using a nationally representative cohort of over 17 000 families within the UK. It adds to and addresses existing literature by considering, in this specific setting, both maternal underweight and overweight, by disentangling the total effect of maternal pre-pregnancy BMI from confounding variables, by looking at the extent of mediation by observed factors, and by considering interactions by ethnicity.
Methods
Sample
The MCS is a nationally representative longitudinal survey of 18 552 families with live babies born in the UK in a period of just over a year, between 2000 and 2002.24 In England and Wales, the period started on 1 September 2000. In Scotland and Northern Ireland, it started on 1 December 2000. Disproportionate sampling ensured adequate representation of the smaller countries and ethnic minority groups,24 stratifying by country and clustering by electoral ward; wards were stratified into ‘advantaged’, ‘disadvantaged’ and ‘ethnic’ areas. Only areas in England were included in the ‘ethnic’ stratum. Six waves of data collection have taken place to date; this study uses the first wave only. Wave 1 interviews took place when babies were around 9 months old.
The analytic sample (n=17 113) includes those who had both information for breastfeeding (n=18 506, no missing data) and maternal pre-pregnancy BMI (n=17 113, 7.53% missing). Fifty-five cases where the main survey respondent is not the baby’s biological mother are removed from the sample. Multiple births (n=224 in the analytical sample) are included. Cases of all gestations are included and preterm delivery (n=1 241) is accounted for in analyses.
Measures
Exposure: maternal pre-pregnancy BMI
Mothers were asked at first interview, ‘how tall are you (without shoes)’ and ‘Thinking back to just before you became pregnant with [baby], what was your weight then (without clothes)?’. Reported pre-pregnancy weight (in kilos) is divided by height (in metres) squared to produce pre-pregnancy BMI. In line with WHO recommendations, BMI is banded into ‘underweight’ (<18.5; 5.84% of the sample); ‘healthy weight’ (18.5–24.99; 65.83%); ‘overweight’ (25–29.99; 19.71%); and ‘obese’ (30+; 8.62%).25 BMI is also used as a continuous variable in an alternative specification (mean=23.64; 95% CI 23.54 to 23.73).
Outcomes: breastfeeding initiation, early cessation and longevity
Mothers were asked at first interview, ‘Did you ever try to breastfeed [baby]?’ and ‘How old was [baby] when [s]he last had breast milk?’ Answers to these questions are combined to produce indicators of whether: (A) breastfeeding was initiated (70.06% of the sample); (B) ceased during the first 7 days, conditional on initiation (17.93% of the sample who initiated); (C) continued for more than 4 months (the point until which exclusive breastfeeding was recommended by the UK’s NHS in 2000–200226), conditional on initiation (39.63% of the sample who initiated). These definitions of breastfeeding behaviours highlight three distinct points in the breastfeeding trajectory, within the context of the mothers’ personal experience and the contemporaneous advice and recommendations to which she was subject. Potentially, different intervention may be feasible at each distinct point, so disambiguating in this way is of interest. Help with intentionally planning and beginning breastfeeding, help with establishing breastfeeding in the early days, and help with continuing to breast feed in the longer term, once it is well established and the infant is introduced to additional nutritional sources, may require separate strategies.
Confounders
Variables indicated as potentially confounding by previous research, as available in the dataset, are treated as such unless otherwise indicated: mother’s ethnicity, mother’s education, mother’s age, siblings in home, presence of step-siblings, smoking during pregnancy, and drinking alcohol during pregnancy (see table 1 for detail).
Mediators
Variables suggested by previous research as potentially mediating the relationship between BMI and breastfeeding are accounted for: birth weight (low: <2.5 kg; normal: 2.5–3.999 kg; heavy: ≥4 kg)), gestational age (pre-37 or post-37 weeks), mother’s gestational diabetes, method of delivery, and labour complications as reported by the mother (as opposed to her reporting she experienced no complications) (see table 1 for detail). Gestational weight gain is not available in the MCS.
Conceptual model
DAGs are used in order explicitly to state model assumptions and reduce bias in estimated outcomes.20 21 Existing evidence, described in the previous section, is used to design the conceptual model, and open source software DAGitty is used to draw the conceptual model and establish appropriate minimal adjustment to block biassing pathways.27 A simplified version of this conceptual model, which does not show inter-relationships between confounder or mediator variables, is presented in figure 1. The full conceptual model is shown in web online supplementary appendix 1.
Supplementary file 1
Statistical analysis
Analyses were conducted using Stata V.14. The complex sampling design of the MCS was accounted for using the ‘svy’ commands in Stata, declaring the strata, clusters, probability weight and finite population correction factor. For the results presented here, the MCS is weighted using the ‘aovwt2’ weight, which adjusts observations to be nationally representative for the whole of the UK based on the probability of selection into the survey, and non-response based on observable characteristics.24 Descriptive statistics calculated unadjusted ORs and 95% CIs for the relationship between pre-pregnancy BMI category and breastfeeding initiation, early cessation and longevity. These unadjusted relationships were estimated separately for each ethnicity.
The total effect of pre-pregnancy BMI category on breastfeeding initiation, early cessation, and longevity, respectively, was estimated by adjusting for confounding variables in logistic regression models. The direct effect of pre-pregnancy BMI category was estimated by further adjusting for mediator variables. Results are presented as ORs, and as predicted marginal probabilities, to assist with the assessment of the relative and absolute size of the effects of pre-pregnancy BMI.
To test whether the effect of pre-pregnancy BMI is moderated by ethnicity, interactions between ethnicity and pre-pregnancy BMI category were added. Sensitivity analyses were conducted using BMI as a continuous variable in initiation, early cessation and longevity models. Where confounders or mediators had missing information, missing categories were included in the analysis such that a common sample was analysed to obtain total and direct effects.
As a further sensitivity analysis, longevity was measured using breastfeeding duration in months with analysis conducted using ordered logistic regression. This analysis is available in web online supplementary appendix 2.
Supplementary file 2
Results
Descriptive
The total analytic sample was 17 113 (92.5%) mothers. The confounders and mediators either had no missing data (ethnicity, siblings in household, step-siblings in household, smoking during pregnancy, gestational diabetes, delivery method and labour complications), or very low percentages of missing data <1% (maternal education, maternal age, alcohol consumption during pregnancy, birth weight and gestational age).
Of this sample, 70% of mothers initiated breastfeeding. Initiation rates varied considerably by ethnicity (white (68.19%), mixed (87.97%), Indian (84.65%), Pakistani (76.81%), Bangladeshi (87.67%), black Caribbean (92.03%), black African (95.31%) and ‘other’ ethnicity (91.37%)). Of those who initiated, 17.93% ceased breastfeeding at 1 week or earlier, and 39.63% breast fed for more than 4 months.
Table 2 shows the unadjusted relationship between pre-pregnancy BMI category and breastfeeding initiation, cessation and longevity for the whole analytic sample and stratified by ethnicity. For the whole sample, being underweight, overweight or obese is associated with decreased odds of initiating breastfeeding (underweight OR=0.63 (95% CI 0.54 to 0.74), overweight OR=0.86 (95% CI 0.77 to 0.96), obese OR=0.83 (95% CI 0.72 to 0.96)), increased odds of ceasing breastfeeding within the first week (underweight OR=1.32 (95% CI 1.05 to 1.66), overweight OR=1.59 (95% CI 1.38 1.83), obese OR=1.63 (95% CI 1.35 1.97)) and decreased odds of breastfeeding for 4 months or more (underweight OR=0.79 (95% CI 0.65 to 0.96), overweight OR=0.75 (95% CI 0.67 to 0.84), obese OR=0.58 (95% CI 0.48 0.69)), compared with those who had a healthy weight prior to pregnancy. The results were very similar for the white ethnic group, who comprise almost 90% of the sample.
Despite the wide CIs, the patterning of the prevalence and the ORs by BMI category suggests potentially varying relationships between pre-pregnancy BMI and breastfeeding behaviours for some ethnic groups. However, formal tests revealed no evidence for multiplicative interaction effects for initiation (F(18,372)=1.07, p=0.39). Small numbers of those classified as underweight in some ethnic groups with very high rates of initiation (mixed (n=11), black African (n=13) and black Caribbean (n=11)) limit the ability of analyses to estimate the impact of underweight on breastfeeding initiation in these groups. There was however evidence for multiplicative interaction effects for early cessation (F(20,370)=2.30, p=0.001) and longevity (F(21,369)=1.91, p=0.01) in unadjusted models.
Total and direct effects
The total effect and direct of pre-pregnancy BMI groups and continuous BMI on breastfeeding initiation, early cessation and longevity is estimated in table 3. After adjusting for confounding variables, there was no difference in the odds of breastfeeding initiation (OR=0.95 (95% CI 0.79 to 1.15)), early cessation (OR=0.96 (95% CI 0.74 to 1.23)) or breastfeeding for at least 4 months (OR=1.14 (95% CI 0.91 to 1.42)) for underweight mothers compared with healthy weight mothers. Overweight mothers had lower odds (OR=0.88 (95% CI 0.78 to 1.00)) of initiating breastfeeding compared with healthy weight mothers. The estimated OR suggests that obese mothers also had lower odds of initiating breastfeeding (OR=0.91 (95% CI 0.77 to 1.07)), but CIs suggest it is a reasonable possibility that there was no difference on average between obese and healthy weight mothers.
Overweight (OR=1.62 (95% CI 1.41 to 1.88)) and obese (OR=1.62 (95% CI 1.33 to 1.97)) mothers had higher odds of ceasing breastfeeding within the first week compared with healthy weight mothers. Furthermore, overweight (OR=0.74 (95% CI 0.66 to 0.83)) and obese (OR=0.57 (95% CI 0.48 to 0.69)) mothers had lower odds of breastfeeding for over 4 months compared with healthy weight mothers. The estimated ORs were largely unchanged after adjustment for mediating variables. This suggests only a very small proportion of the total effect is mediated by the available selected variables.
Interactions between ethnicity and BMI group were tested for all three outcomes in total and direct effect models. The results were consistent; here we present the values from direct effect models. The interaction between ethnicity and BMI group was not significant for initiation: (F(18,372)=1.11, p=0.34) or longevity: (F(21,369)=1.37, p=0.13). However, a significant interaction between BMI group and ethnicity for early cessation (F(20,370)=2.10, p=0.004)) suggests the effect of pre-pregnancy BMI on early cessation differs across ethnic groups. For white and Indian mothers being overweight or obese was associated with an increased probability of ceasing breastfeeding in the first week. For Pakistani mothers, being obese pre-pregnancy was associated with a lower probability of ceasing breastfeeding in the first week. However, the number of mothers who cease breastfeeding in the first week, who are underweight or obese and belong to any ethnic group except white, is very small. Therefore, this interaction needs to be interpreted cautiously and is likely very sensitive to small changes in the sample. Furthermore, when overweight and obese groups are combined, the interaction is no longer significant (F(13,377)=1.46, p=0.13)
The marginal probabilities shown in table 3 put the ORs into perspective: holding both cofounder and mediator variables constant across BMI groups, there is only a small (two percentage points) difference in predicted initiation rates for overweight (68.58%) and healthy weight (70.82%) mothers. There is, however, a larger difference in breastfeeding for 4 months or more for healthy weight (38.04%), overweight (32.11%) and obese (28.00%) mothers and also in early cessation rates for healthy weight (18.35%), overweight (25.86%) and obese (25.45%) mothers.
Sensitivity analysis: BMI as a continuous predictor
As shown in the lower part of table 3, after accounting for confounders and mediators, a one-unit increase in pre-pregnancy BMI was associated with 0.99 (95% CI 0.98 to 1.00) times the odds of initiating breastfeeding, 0.95 (95% CI 0.94 to 0.96) times the odds of breastfeeding for at least 4 months and 1.04 (95% CI 1.03 to 1.06) times the odds of ceasing breastfeeding in the first week. There was no evidence for an interaction between ethnicity and BMI in initiation (F(7,383)=0.73, p=0.64), early cessation (F(7,383)=1.64, p=0.12) or breastfeeding for at least 4 months (F(7,383)=0.77, p=0.62).
Discussion
Congruent with previous research from non-UK countries,8–12 this analysis finds unadjusted associations between pre-pregnancy BMI and breastfeeding initiation, early cessation and longevity. However, in contrast to emerging international evidence, after adjustment for confounders, there was no relationship between pre-pregnancy underweight and breastfeeding behaviours. The relationships between overweight and obesity and breastfeeding early cessation and longevity hold as both statistically and substantively important after accounting for a variety of potential confounders and mediators. Pre-pregnancy BMI appears, therefore, to be a descriptively useful predictor of breastfeeding duration. At best, it potentially holds some explanatory power in elucidating the tendency of some mothers in the UK to cease breastfeeding well before the duration recommended by the World Health Organisation.28
The reasons for differences in first-week cessation and longevity beyond the recommended (at the time) duration for exclusive breastfeeding, according to pre-pregnancy overweight, require further exploration. The combined influence of the five mediator variables tested did not explain a substantive proportion of the total effect. Previous studies indicate, tentatively, that psychological factors such as body image may play a part in breastfeeding behaviours.29 Physiological explanations have also been proposed, such as dysregulation of the hypothalamic-pituitary-gonadal axis and in fat metabolism affecting milk production and composition.30 Low prolactin levels in response to infant suckling after initiating breastfeeding reducing the ability of overweight/obese women to produce milk is another possible explanation,31 along with delayed onset of lactation perhaps resulting in early cessation.32 33
Whatever the explanation for the associations, pre-pregnancy overweight and obesity may provide a useful means through which to target early postnatal support services, which have been indicated as effective in promoting initiation and continuation of breastfeeding.34–36 BMI at first antenatal appointment is easily and reliably measured. By prioritising overweight and obese mothers for breastfeeding support, both in the initial period of establishment and in later months, services may reach those less likely to continue breastfeeding. This may prove an efficient use of resources: both because of the estimated health benefits to breastfed babies and their mothers and because research suggests relationships between maternal prenatal BMI and child BMI and cognitive development—associations within which breastfeeding may be a malleable mediating factor.
Strengths and limitations
This study provides evidence on the relationship between pre-pregnancy BMI and breastfeeding behaviours using rich data from a recent, nationally representative sample of UK families. It contributes to the literature by providing evidence on the relationships between pre-pregnancy underweight, as well as overweight, and breastfeeding behaviours, as well as testing whether relationships are moderated by ethnicity. It also accounts for confounding of associations by pre-pregnancy factors, as well as testing channelling of possible effects through mediators occurring during gestation and birth. Unfortunately, however, the data in the MCS does not allow control for gestational weight gain, which may be related to pre-pregnancy BMI and therefore is a potential untested mediator.
A further limitation to this study is the reliance on self-reported height and weight and the recording of pre-pregnancy weight retrospectively. Evidence suggests self-reported anthropometric measures are valid despite a tendency for height to be over-reported and weight to be under-reported.37 Moreover, there is a relatively short time period of recall in the MCS, which suggests reasonably accurate estimates of weight.38 However, comparisons of reported and measured height and weight demonstrate an over-reporting of underweight status and an under-reporting of obese status.39 Evidence also suggests that for overweight and obese mothers, pre-pregnancy weight is under-reported to a greater extent than for healthy weight mothers.40 41 Estimates of the relationship between pre-pregnancy overweight and obesity may therefore be biased downwards, as some women who are in reality obese/overweight may be classified as overweight/healthy weight in our analysis, which means the effects are underestimated. These underestimates should not affect the overall conclusions drawn from these analyses.
The advantage of choosing BMI as a measure here is that it is widely used, making results easily applicable and interpretable. The sensitivity check where BMI is used as a continuous predictor rather than a grouped variable alleviates the possibility that results may be an artefact of imposed cut-points.
Binary logistic regressions are used for all analyses in this paper, as our outcomes are dichotomous. The resulting odds ratios can at times be misleadingly interpreted – to mitigate this, and for clarity,model-estimated marginal means are presented in each table of results. These estimates represent the percentage probability for each group for each outcome, thus conveying meaningful effect sizes.
Conclusion
Further research is needed into the causal mechanisms underlying the relationship between maternal pre-pregnancy overweight and obesity and breastfeeding behaviours, particularly given tentative indications in analyses here that there may be variations by ethnic group in the UK context. Scant explanation of the tendency to lower breastfeeding rates among overweight and obese mothers is offered by the pregnancy and delivery factors explored here. Nevertheless, this lack of attenuation may be pragmatically useful. This research suggests that pre-pregnancy BMI could be one predictive measure in targeting of support to some of the women who are substantially less likely to establish breastfeeding or to continue in the longer term. The nature of this support should carefully be considered and tailored according to further theory and research, in order to maximise benefits for babies and mothers and minimise the risk of unintended negative effects of intervention.
What is already known on this subject
Breastfeeding is beneficial to child health and development, with current recommendations of exclusive breastfeeding for the first 6 months of life. However, international literature suggests that rates of breastfeeding initiation and duration vary by maternal pre-pregnancy body mass index (BMI).
What this study adds
In the UK, there are large differences in breastfeeding early cessation (stopping within first week) and longevity (breastfeeding for over 4 months) by maternal pre-pregnancy BMI group. Though they are similarly likely to initiate, overweight and obese mothers are less likely to establish breastfeeding in the early days and less likely to continue in the longer term.
Alongside notable differences in breastfeeding behaviours across mothers’ ethnic groupings, there is some very tentative evidence that the relationship between maternal BMI groupings and early cessation of breastfeeding varies by ethnic group, although this does not appear to be a robust association within this sample.
Other candidate mediators which might suggest mechanisms in the BMI – breastfeeding relationship, including gestational diabetes, gestational age, birthweight, delivery method and whether there were self-reported labour complications, were tested. They do not attenuate or explain the relationship between pre-pregnancy overweight and obesity and breastfeeding early cessation or longevity.
This research suggests that pre-pregnancy BMI could be one predictive measure in targeting of support to women substantially less likely to breast feed for the recommended duration. However, the nature of support should carefully be tailored according to further research on explanations and mechanisms, and trialled carefully and mindfully with regard both to intended and possible unintended consequences.
Acknowledgments
This paper uses data from the Millennium Cohort Study. Many thanks to all the study participants. The authors are also grateful to the Centre for Longitudinal Studies, Institute of Education, for the use of these data, and to the UK Data Archive and Economic and Social Data Service for making them available. However, these organisations bear no responsibility for the analysis or interpretation of these data. Data citation: University of London. Institute of Education. Centre for Longitudinal Studies, Millennium Cohort Study: First Survey, 2001-2003 (computer file). 11th Edition. Colchester, Essex: UK Data Archive (distributor), December 2012. SN: 4683, http://dx.doi.org/10.5255/UKDA-SN-4683-3
References
Footnotes
TC and NS contributed equally.
Contributors Both authors contributed equally to this paper.
Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.
Disclaimer All views and interpretations expressed in this article are the authors’ own.
Competing interests None declared.
Patient consent Not required.
Ethics approval Because this study is a secondary data analysis, research ethics approval was not required. The first wave of the Millennium Cohort Study received ethics approval from National Health Service Ethical Authority in February 2001 (MREC/01/6/19).
Provenance and peer review Not commissioned; externally peer reviewed.
Data sharing statement Not applicable; this is secondary data.