Introduction

The number of children in Scotland with a record of special educational need (SEN) increased almost 4-fold between 2010 and 2018 [1], and, by 2020, almost a third of pupils in Scotland had a record of SEN [2]. While higher case ascertainment may have contributed, it nonetheless represents a significant burden on the education, health, and social sectors, as well as substantial impact on the affected individuals, their families, and wider society [3]. Children with SEN experience lower educational attainment [4,5], higher rates of school absenteeism and exclusion, and higher rates of bullying and maltreatment [6–8], further impacting their physical and mental health and well-being [9]. Discovering modifiable early life risk factors for SEN is therefore important to enable these burdens to potentially be eased through prevention or earlier detection.

Method of infant feeding is one factor that is worth investigating when considering early life risk factors for SEN. It is the recommendation of the World Health Organisation (WHO) that, for the first 6 months of their life, children be breastfed exclusively, due to the health benefits that it affords to both mother and baby [10,11]. Breastfeeding has been shown to be associated with reduced risk of a range of physical health outcomes including gastrointestinal, respiratory, and urinary tract infections, otitis media, asthma, obesity, and diabetes [12,13]. There is also evidence of reduced risk of conditions known to be associated with SEN: autistic spectrum disorder (ASD) [14,15], attention deficit hyperactivity disorder (ADHD) [16,17], communication problems [18,19], mental health problems [20,21], behavioural problems, and impaired social development [22,23], and evidence that breastfeeding may be associated with higher levels of intelligence in later life [24,25]. There is therefore the possibility that breastfeeding could be protective against the development of learning problems, and, given that the most common reason for SEN in Scotland is learning difficulties or disabilities [26], this adds further strength to the hypothesis that infant feeding method may influence the development of all-cause SEN. The nutritional, hormonal, and chemical components of breast milk may directly impact neurodevelopment and risk of disease [27–29]. However, the maternal–infant bonding associated with breastfeeding may also affect child development [30,31].

Previous studies on the associations between infant feeding method and overall risk of outcomes that can contribute to SEN have been limited in number, unable to adjust comprehensively for child, maternal, and pregnancy confounders [32–34], prone to selection bias or recall bias [35,36], unable to take account of mixed feeding [8,32–39], or used a narrow definition of SEN covering only intellectual disability [35]. This study aims to add to the existing evidence and address these limitations by linking national, routinely collected health and educational data together to investigate the association between mode of infant feeding at 6 to 8 weeks of age and both all-cause and cause-specific SEN in Scotland. To the best of our knowledge, the associations between infant feeding method and risk of outcomes contributing to SEN have not previously been investigated on a population-wide scale. Furthermore, this study fills a gap in the literature by investigating a wider range of outcomes than previous studies, investigating formal school-recorded data on SEN, and comparing exclusively breastfed children against mixed-fed children and formula-fed children. We hypothesise that children who are exclusively breastfed will have lower risk of subsequent all-cause and cause-specific SEN compared to children who are mixed-fed and formula-fed.

Methods

This study is reported as per the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guideline (S1 STROBE Checklist). While we did not publish an analysis plan, our analyses were planned before the research team accessed any data. Data were extracted in June 2015, as part of a wider linkage project [40], and this specific study commenced in February 2021, utilising the same data.

Results

The original cohort after data cleaning comprised of 2,793,185 education records pertaining to 766,244 singleton school children born in Scotland and attending school in Scotland between 2009 and 2013. This reduced to 694,875 records pertaining to 238,171 children after limiting to children born on or after 2004. It reduced further to 555,593 records pertaining to 191,745 school children who had complete feeding information recorded at 6 to 8 weeks of age (Fig 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Flow diagram presenting the number of records and pupils excluded at each stage of data cleaning.

CHI, Community Health Index; P1–P7 primary 1 through primary 7; SMR, Scottish Morbidity Record; S1–S6, secondary 1 through secondary 6.

https://doi.org/10.1371/journal.pmed.1004191.g001

Of these 191,745 children, 126,907 (66.2%) were formula-fed, 48,473 (25.3%) were exclusively breastfed, and 16,365 (8.5%) were mixed-fed. All those children, regardless of age, had breastfeeding data, maternity data, and SEN data available. Breastfed children were more likely to be female, lived in more affluent areas, and had mothers who were older, less likely to smoke, and more likely to be married and parous (Table 1). They were less likely to be small for their gestational age, and more likely to be full term at delivery, have had a cephalic vaginal delivery and a 5-minute Apgar score of 7 to 10. The variables with the greatest amount of missing data were marital status (39.93%), smoking during pregnancy (13.99%), and pupil ethnicity (3.45%). However, missing values for these 3 variables were analysed as “unknown” categories and included in all analyses to minimise loss of records. The remaining variables were ordinal categories; therefore, including missing data as unknown categories did not make sense. However, these variables had lower levels of missing values: 5-minute Apgar score (2.81%), child deprivation quintile (2.53%), and parity (1.01%), with the remainder all less than 0.2%. Therefore, we did not deem multiple imputation to be necessary and instead used complete case analyses.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Child, maternal, and pregnancy characteristics by feeding method at 6–8 weeks of age.

https://doi.org/10.1371/journal.pmed.1004191.t001

Overall, 23,141 (12.1%) children had a record of SEN: 8,878 (4.6%) had learning difficulties, 6,022 (3.1%) social emotional behavioural difficulties, 5,629 (2.9%) communication problems, 4,389 (2.3%) learning disabilities, 2,137 (1.1%) ASD, 1,749 (0.9%) physical motor impairments, 1,579 (0.8%) physical health problems, 1,223 (0.6%) sensory impairments, and 157 (0.1%) mental health problems. Numbers and percentages of these outcomes by specific feeding type are presented in Table 2.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Proportion of children with all-cause SEN and cause-specific SEN by feeding method at 6–8 weeks of age.

https://doi.org/10.1371/journal.pmed.1004191.t002

Compared with children who were formula-fed, we observed that children who were mixed-fed (OR 0.70, 95% CI [0.66,0.74], p < 0.001) and children who were exclusively breastfed (OR 0.59 (95% CI [0.57,0.61], p < 0.001) were both less likely to have all-cause SEN on univariate analysis. Following adjustment for child, maternal, and pregnancy confounders, the associations were attenuated but remained significant (mixed feeding OR 0.90, 95% CI [0.84,0.95], p < 0.001 and exclusive breast-feeding OR 0.78, 95% CI [0.75,0.82], p < 0001) (Table 3).

Compared with children who were formula-fed, we observed that children who were mixed-fed or were exclusively breastfed were less likely to have learning disabilities (mixed feeding OR 0.75, 95% CI [0.65,0.87], p < 0.001 and exclusive breastfeeding OR 0.66, 95% CI [0.59,0.74], p < 0.001), or learning difficulties (mixed feeding OR 0.85, 95% CI [0.77,0.94], p = 0.001 and exclusive breastfeeding OR 0.75, 95% CI [0.70,0.81], p < 0.001). Compared with children who were formula-fed, we observed that children who were exclusively breastfed were less likely to have communication problems (OR 0.81, 95% CI [0.74,0.88], p < 0.001), social–emotional–behavioural difficulties (OR 0.77, 95% CI [0.70,0.84], p < 0.001), sensory impairments (OR 0.79, 95% CI [0.65,0.95], p = 0.01), physical motor disabilities (OR 0.78, 95% CI [0.66,0.91], p = 0.002), and physical health conditions (OR 0.74, 95% CI [0.63,0.87], p < 0.001) after adjusting for potential confounders. Feeding method was not significantly associated with ASD or mental health conditions (Table 3).

No statistically significant interactions between feeding method and either pupil sex, pupil age, or pupil socioeconomic status were identified. Furthermore, exclusion of the 7,533 children who had a congenital abnormality and exclusion of the 13,193 children who were admitted to intensive care or special care baby units did not alter the findings (S1 Table).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Univariate and multivariate associations between infant feeding method at 6–8 weeks of age and all-cause and cause-specific SEN.

https://doi.org/10.1371/journal.pmed.1004191.t003

Discussion

In this study, we observed that compared to children who were formula-fed, exclusively breastfed children had less risk of all-cause SEN and SEN attributed to learning disabilities, learning difficulties, communication problems, social–emotional–behavioural difficulties, sensory impairments, physical motor disabilities, and physical health conditions after adjusting for available sociodemographic and maternity factors. Compared to children who were formula-fed, those who were mixed-fed also appeared to be less likely to have all cause SEN and SEN attributed to learning disabilities or learning difficulties.

Our findings concur with a previous case–control study in which 49 individuals aged 13 to 22 years who were receiving additional learning support for presumed cognitive difficulties were less likely to have been breastfed on discharge from hospital than their siblings and an unrelated control group [35]. Our finding that breastfeeding was associated with less communication problems is consistent with the findings in previous studies [18,19]. Social–emotional–behavioural problems have also been examined previously, but with both exclusive and nonexclusive breastfeeding being found to be protective, in contrast to our findings [17,22]. Surprisingly, our study found no significant relationship between either breastfeeding or mixed feeding and ASD. This conflicts with two previous meta-analyses that both reported that breastfeeding was protective against ASD [14,15]. This may be due to our study including less severe cases in which the association may be weaker or nonexistent.

Congenital anomalies can impact on infant feeding choices [47–49], as well as predisposing to SEN [36,50]. Similarly, admission to intensive care and special care baby units can impact feeding choice [51–53] and is associated with SEN [35]. Many previous studies have included children with congenital anomalies or those admitted into neonatal units but not adjusted for these as potential confounders [8,32,33,36,38,39]. In our study, we observed that the association between breastfeeding and reduced risk of SEN was present among children who did not have congenital anomalies and among children not admitted to intensive care or special care baby units. However, a limitation of our study was lack of data on specific lengths of stay in intensive care or special care baby units.

Children who were born in private hospitals, privately educated, or homeschooled were not included in the study. However, only 4% of children in Scotland attend an independent school [54], and homeschooling and private maternity care are very uncommon [55]. Because of the need to link birth and education records, we could not include children who were not born in Scotland or who emigrated from Scotland before starting school.

Previous studies on breastfeeding and SEN have analysed smaller study populations [35,36] and been prone to selection bias [8,38], recall bias [8,33,34,36,38], and loss to follow-up. Use of national routine data enabled us to conduct a large-scale, unselective study that included hard-to-reach groups. Since both exposure and outcome data were collected routinely by the health and education sectors, recall bias was avoided.

Many previous studies have analysed feeding method as a binary variable [32,33,35]. We were additionally able to investigate children who were mixed-fed. While self-reported feeding method was based on predominant feeding method in the preceding 24 hours and was therefore not affected by recall bias, it was not corroborated and therefore could not differentiate infants who had never been breastfed from those in whom it had been discontinued prior to 6 weeks. Scottish data show that rates of breastfeeding drop off between first visit, 6- to 8-week review, and 13- to 15-month review [56]. Therefore, status at 6 to 8 weeks gives a snapshot picture but cannot be applied to wider time points. We also did not have information on the type of formula feed, whether any children received donor milk, or use of expressed breast milk. The latter is important since skin-to-skin contact increases maternal bonding and is beneficial for cognition [30,31].

Health visitors are funded by the National Health Service, which is free to the whole population at the point of delivery. The Scottish Government’s core programme of child health reviews and screening activities are a universal process that are offered to every child. Beyond core contacts, the health visitor uses professional judgement to decide on further contacts. While there is no legal requirement to have a child health review, and it is not mandatory that mothers accept a child health visit, recent data from Public Health Scotland show that 97% of children successfully receive their first child review (in the days after birth) and 90% successfully receive their 6- to 8-week review with uptake similar across different ethnic groups, maternal ages, and deprivation categories. [57] In our study, based on older data, around 20% of pupil records had no 6- to 8-week review data; however, this figure was similar regardless of mother’s maternal age, or pupil’s ethnicity or deprivation, suggesting that missing data on more vulnerable groups of children is unlikely. Indeed, the percentage of missing data was least among most deprived children and children of Asian ethnicity.

The risk of incomplete or inaccurate data inherent in secondary data analysis was mitigated by the recording of SEN being a statutory requirement. Prevalence of SEN increased over the duration of the study from 2% to 12% and has continued to increase in recent years. This increase can be largely attributed to increased ascertainment of conditions such as ASD and ADHD due to greater awareness among parents and teachers. Since the observed increase was similar irrespective of exposure group, it is unlikely to have introduced bias. We also adjusted all of our analyses for pupil age. The requirement for SEN is decided by the teacher, and the school’s special educational needs coordinator (SENCO) in discussion with the child’s parents or guardians, taking account of formal and informal assessments and wider information about progress and desired outcomes. The additional support put in place can vary significantly and is dependent on the individual child’s requirements [3]. The categories for the cause-specific SENs are completed within the education sector, and the person deciding on what is recorded can vary from teacher to medical professional to support staff such as an educational psychologist. It is possible that some children with needs will not meet the threshold to be formally offered support while others may have difficulties that are not picked up. The study was retrospective and used administrative data. Therefore, it is not possible to verify that ascertainment of SEN was 100% complete as this would require prospective recruitment and examination/investigation of all children. However, likelihood of complete ascertainment is increased due to the benefits of formal diagnoses, the legal requirements to identify and record SEN, and the fact that ascertainment can be triggered by teachers, parents, and healthcare workers. Furthermore, there is no reason to believe that ascertainment is likely to differ systematically by breastfeeding history, and, therefore, that bias was introduced.

The results of this study were obtained based on Scottish data. Rates of breastfeeding, however, vary greatly across the United Kingdom, Europe, and worldwide. It has been reported that overall prevalence of breastfeeding varies from 30.7% in high-income countries to over 90% in low- and middle-income countries [58]. This makes it difficult to generalize our findings to low- and middle-income countries. Our observed breastfeeding rate of 33.8% (either exclusive or mixed breastfeeding) agrees broadly with rates reported across high-income countries.

We had access to several potential confounders not available in some previous studies [8,35–37,39], but, as with any observational study, residual confounding is possible. We did not have any data on important maternal or paternal factors such as education level [59,60], intelligence quotient (IQ) [61], employment status, race/ethnicity, or mental and physical health; however, we did adjust for the mother’s area socioeconomic status at the time of birth, which is a proxy measure. One of the most frequently seen confounders within the literature is parental education or intelligence level. Children born to mothers or fathers who are more educated or have a higher IQ appear to have lower likelihood of the most common cause of SEN—learning difficulties [59,60,62]. There is also evidence that parents who have reached higher education levels are more likely to breastfeed [63,64]. Other sociodemographic factors may also act as confounders in the relationship. Older mothers, those from Black and Minority ethnic groups, those who are married or cohabiting, and those in higher socioeconomic classes have been reported to be more likely to breastfeed [63,65]. Some of these characteristics are also associated with intellectual problems. For example, it has been stated that divorced or single mothers are more likely to have children with cognitive delay [66]; children from families with a higher socioeconomic position are more likely to do better academically [60,67]; and, among late and moderately preterm children, non-white ethnicity is a significant risk factor for cognitive impairment [35]. Our data agreed with previous literature with respect to breastfeeding rates being lower among younger and more deprived mothers. The reasons for these patterns should be explored in more detail.

There are several areas where further research would be beneficial. Firstly, being able to examine the impact of duration of different feeding types on all-cause and cause-specific SEN in a large population cohort such as this would provide insights into any dose–response relationships that may be present, and whether there are any critical feeding durations. Furthermore, research regarding expressed breast milk versus breastfeeding on a large representative sample would enhance understanding of whether it is the act of breastfeeding, or breast milk itself that impacts upon SEN. Finally, with regard to the exposure, it would be useful to know whether the relationship between feeding and SEN is maintained for different types of formula feed, as well as donor milk compared with mother’s milk.

Residual confounding could be responsible for the observed associations in this study, and so performing the same methodology on a large sample with a greater number of confounders is also recommended, particularly including confounders that were not included in this study due to lack of data such as maternal IQ, parental education level and occupation, and alcohol or drug use during pregnancy.

Future studies are also required to investigate whether the impact of neonatal stays on the relationship between mixed feeding and SEN applies for all lengths of neonatal stay. Unfortunately, this was not possible to conduct within this piece of research due to lack of available data around length of stay. Finally, there has been some research that suggests that breastfeeding is linked to reduced hospitalisation and reduced chronic disease in childhood [12,68–70]. Certain chronic diseases have also been associated with SEN [71–74]. There is the possibility, therefore, that childhood chronic disease acts as a mediator between feeding method in infancy and SEN, and so further research is required to examine this relationship.

In conclusion, breastfeeding may be protective against children developing SEN. While exclusive breastfeeding is the ideal, even mixed feeding may confer some benefit. The results of this study suggest that feeding method in infancy could be a modifiable risk factor for all-cause SEN, which has the potential to help reduce its burden in relation to the affected children, their families, and wider society. We observed that, compared to formula-fed children, children who were breastfed and children who were mixed-fed at 6 to 8 weeks had lower risk of all-cause SEN, and SEN attributed to learning disabilities and learning difficulty. This is significant as many women struggle to exclusively breastfeed for the full 6 months recommended by WHO [10] for a host of different reasons [11]. However, this study provides evidence that even a short duration of nonexclusive breastfeeding could be beneficial with regard to development of SEN. Our findings augment the existing evidence base concerning the advantages of breastfeeding and reinforce the importance of breastfeeding education and support. Several meta-analyses have been conducted examining the effectiveness of breastfeeding education and support programmes, and these all concluded that the programmes were beneficial as regards increasing the initiation and duration of breastfeeding [75,76]. Despite breastfeeding rates increasing in recent years, the rates in Scotland vary by maternal age, socioeconomic status, and education level and are lower than rates in England [56]. Policies such as Scotland’s 2019 published recommendations to make the country breastfeeding friendly [77] should therefore remain at the forefront of political and public health agendas.

Supporting information