G Model
JCD-5703; No. of Pages 12 Journal of Communication Disorders xxx (2015) xxx–xxx
Contents lists available at ScienceDirect
Journal of Communication Disorders
Review
Breastfeeding and language outcomes: A review of the literature [9_TD$IF]J. Mahurin[10_TD$IF]-Smith * Illinois State University, Normal, IL 61790, United States
A R T I C L E I N F O
A B S T R A C T
Article history: Received 18 September 2014 Received in revised form 10 April 2015 Accepted 19 April 2015 Available online xxx
Many researchers have investigated the potential impact of breastfeeding in infancy on a child’s subsequent development, but only a small subset of these studies considers language development and impairment. This paper reviews that literature, discussing postnatal neurodevelopment, potential mechanisms for dietary influences on communication outcomes, studies of typically developing children, and studies of children with communication concerns. For population based studies of language development, a modest but statistically robust relationship is seen across large samples that account for breastfeeding exclusivity. A similar protective relationship is seen in studies that evaluate the relationship between breastfeeding and language disorders; effect sizes are typically larger in these papers. Implications for researchers and service providers are reviewed. Learning outcomes: Readers will be able to describe possible mechanisms by which early diet might influence neurodevelopment. They will be able to describe the relationships observed between diet in infancy and language outcomes in large population-based studies, as well as the trends observed in studies of the relationship between infant diet and communication impairment. ß 2015 Elsevier Inc. All rights reserved.
Keywords: Feeding Breastfeeding Child language Language delay Language disorder Neurodevelopment
Contents 1. 2. 3. 4. 5. 6. 7.
8.
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Postnatal neurodevelopment. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Why would diet influence language development? Potential mechanisms . . . . . . . . . . The existing literature on typical development . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . The existing literature on communication disorders . . . . . . . . . . . . . . . . . . . . . . . . . . . . Human neurodevelopment and infant diet: highly publicized null findings . . . . . . . . . Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7.1. Review of findings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7.2. Assessment of the existing research . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7.3. Implications for professionals in communication sciences and disorders (CSD) . Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
000 000 000 000 000 000 000 000 000 000 000 000 000
* Tel.: +1 309 438 5308. E-mail address: [email protected] http://dx.doi.org/10.1016/j.jcomdis.2015.04.002 0021-9924/ß 2015 Elsevier Inc. All rights reserved.
Please cite this article in press as: Mahurin-Smith, J. Breastfeeding and language outcomes: A review of the literature. Journal of Communication Disorders (2015), http://dx.doi.org/10.1016/j.jcomdis.2015.04.002
G Model
JCD-5703; No. of Pages 12 2
J. Mahurin-Smith / Journal of Communication Disorders xxx (2015) xxx–xxx
1. Introduction Imagine that two communities in two different parts of the world are building a network of roads. As a cost-saving measure, the planning teams opt to use lower-grade building materials for half of each community. One community’s climate is mild: temperatures tend to stay in the 40–70 8F range. The second community experiences much more temperature variability: winter temperatures might plummet to 20 8F or cycle rapidly from freeze to thaw and back again, while summer temperatures might hover around 100 8F. Drivers who have navigated pothole-ridden streets in spring might guess that the roads built with lower-grade materials could show signs of wear more rapidly in Community #2. When the strains on the materials are greater, the breakdowns in the network may be apparent more rapidly. Drivers in Community #1 might notice only minor signs of wear on the lowergrade sections of their roadways, while drivers in Community #2 might find a conspicuous difference. The neurological system can be envisioned as a vast network of roads, undergoing large-scale construction projects during pregnancy, infancy, and toddlerhood. The ‘‘road-building’’ materials available to children can vary with their early diet, leading to differences in brain structure (Farquharson, Jamieson, Logan, Cockburn, & Ainslie Patrick, 1992; Farquharson et al., 1995; Jamieson et al., 1999; Makrides, Neumann, Byard, Simmer, & Gibson, 1994). These differences in brain structure may be associated with differences in brain function (Greiner, Moriguchi, Hutton, Slotnick, & Salem, 1999; Salem et al., 2000). For children whose network of genetic and other environmental effects does not predispose them to language difficulties – similar to the roads in Community #1 – diet in infancy may be linked to modest differences in later life (Anderson, Johnstone, & Remley, 1999). For children whose genotype renders them more vulnerable to language impairment, like the roads in Community #2, early diet may make a more significant difference as they mature (Schultz et al., 2006). Breastfeeding is widely recommended for infants, and the differences in morbidity between breastfed and formula-fed children are well-documented (American Academy of Pediatrics, 2012; World Health Organization, 2003). Less familiar, perhaps, are the differences in neurodevelopment that have also been observed in studies of breastfed versus formula-fed children. Although a subset of this literature looks explicitly at the relationship between feeding mode in infancy and subsequent speech-language development, these investigations are seldom referenced in communication sciences and disorders (CSD) journals (Rogers et al., 2015). This paper will first describe the process of postnatal brain development and the mechanisms through which diet might influence neurodevelopment, followed by a review of the existing literature on associations between breastfeeding and subsequent language development and impairment. It will conclude with a review of potential implications arising from these findings. 2. Postnatal neurodevelopment Lawrence and Lawrence (2005) report that the average newborn will arrive with a brain weighing approximately 350 g; a year later, that brain will have grown to approximately 1100 g. By age 3 the brain will have quadrupled relative to its size at birth; the steepness of this early trajectory is particularly clear when one considers that the brain will require a further 15 years to approach the quintuple mark (Dekaban & Sadowsky, 1978). Neurons proliferate in regions including the dentate gyrus of the hippocampus, a region associated with memory; during the first 2 weeks of life this process is especially vulnerable to environmental influences such as the presence or absence of growth factors (Watson, DeSesso, Hurtt, & Cappon, 2006). In addition to the rapid growth of the brain itself, other types of neural tissue proliferate during the early years of life. The most rapid period of postnatal myelination occurs during the first 2 years of life; the major fiber tracts are clearly visible in 3-year-old children (Matsuzawa et al., 2001). During the first 18 months of life neurons must migrate to other parts of the brain, including the prefrontal cortex (Sanai et al., 2011). This is also a period of dramatic growth in synapse formation, so much so that at the point of peak synaptic density there are 55 synapses per 100 mm3 in the human brain (Watson et al., 2006). The decline in synaptic density observed between toddlerhood and adulthood is the result of a form of programmed cell death known as pruning. It is common to think of brain cell death as undesirable, but pruning is a critical part of normal neurodevelopment, distinct from other causes of cell death such as injury or toxin exposure (Watson et al., 2006). Inadequate pruning has been posited as a factor in autism (Hill & Frith, 2003), highlighting the potential importance of this neurodevelopmental process. During infancy the brain is also building its own protective mechanisms, a process that takes time to complete. In the first 6 months of life the blood-brain barrier is more permeable than it will be in later life, increasing the infant’s vulnerability to neurotoxins (e.g., methylmercury) with the potential to affect future cognitive skills (see Watson et al., 2006; Dzwilewski et al., in this issue). To sum up, then, infants not only must build brain cells, they must also establish connections between them, re-sculpt brain architecture, and create a selectively permeable wall between the brain and the outside world. Their diet provides the raw materials available for these tasks. 3. Why would diet influence language development? Potential mechanisms The idea that breastfeeding or formula-feeding could influence language development immediately raises the question of mechanism: why would children’s diets in infancy exert any long-term influence over their speech-language skills? Four possibilities will be reviewed in the section that follows. The explanation most often proposed is that differences in the fatty
Please cite this article in press as: Mahurin-Smith, J. Breastfeeding and language outcomes: A review of the literature. Journal of Communication Disorders (2015), http://dx.doi.org/10.1016/j.jcomdis.2015.04.002
G Model
JCD-5703; No. of Pages 12 J. Mahurin-Smith / Journal of Communication Disorders xxx (2015) xxx–xxx
3
acid profiles of human milk and infant formula might make a difference in subsequent brain structure and function; a closely related possibility is that the fatty acid content of cell membranes could influence gene expression within those cells. The second idea is that human milk might contain a variety of constituents that promote optimal neurodevelopment, perhaps with synergistic effects seen across multiple nutrients. A third consideration is that the well-documented relationship between breastfeeding and immune system function/regulation of inflammation might exert a moderating influence on learning and memory. A fourth potential mechanism involves the impact of lactation on mothers, who in turn shape their children’s learning and language. These mechanisms are not mutually exclusive, and the reader is reminded that a single causal mechanism is unlikely to provide an adequate explanation for a phenomenon as complex as neurodevelopment. Analysis of the brain tissue acquired during infancy indicates that more than half of its solid weight is lipid (Lawrence & Lawrence, 2005), and that the fatty acid profile of an infant’s diet plays a significant role in the composition of brain tissue (Farquharson et al., 1992, 1995; Jamieson et al., 1999; Makrides et al., 1994). Of particular importance are two long-chain polyunsaturated fatty acids: an omega-3 (or n-3) fatty acid known as docosahexaenoic acid (DHA) and an omega-6 (n-6) fatty acid called arachidonic acid (AA). Across mammal species, DHA is the most abundant fatty acid found in the brain, but the amounts of DHA in brain tissue are directly related to an individual’s diet (Innis, 2007). Not all of the infant formulas for sale in the US contain DHA and AA; those that do are consistently more expensive than formulations with supplemental LC-PUFA. For this reason, formula-fed infants in the US may lack a dietary source of DHA and AA. Much of the evidence supporting the relationships among dietary fatty acid intake, brain composition, and learning/ behavioral outcomes comes from animal studies, in which diet can be manipulated to a degree that would be gravely unethical in studies of human infants. It has long been documented that rat pups fed a diet free of n-3 fatty acids build brain tissue with much less DHA and a corresponding increase in n-6 fatty acids (Bourre et al., 1984). These differences in n-3 intake are associated with poorer learning and memory impairment in adult rats (Greiner et al., 1999), as well as worse scores on measures of depression and aggression (DeMar et al., 2006). Studies of human children also suggest that the fatty acid content of the diet influences the structure of neural tissue, both gray matter and white matter. Kafouri et al. (2012) investigated cortical thickness in 571 adolescents and found that increased duration of exclusive breastfeeding was strongly correlated with both increased thickness of the parietal cortex and with higher cognitive scores; the authors posited that the DHA in human milk may explain these findings. Another recent study (Deoni et al., [12_TD$IF]2013) found slower myelination in formula-fed children as compared to their breastfed counterparts. These authors state, ‘‘While the exact mechanism(s) that underlie these. . .differences remain unclear, our results lend support to the hypothesis that the docosahexaenoic and arachidonic acids present in breast milk promote preferential neural growth and white matter development’’ (Deoni et al., 2013, p. 83). Gross quantitative measures of neural tissue offer only a rough index of brain maturation. At a microscopic level, LC-PUFA intake influences the growth of dendritic spines and the creation of synaptic membranes, thus shaping neurotransmission and cell-to-cell signaling (Nyaradi, Li, Hickling, Foster, & Oddy, 2013). For scholars of genetically mediated phenomena such as speech and language, it is perhaps even more important to know that the fatty acid content of cell membranes can shape gene transcription and expression (Innis, 2007). Animal studies have indicated that dietary intake is well-documented as an agent of DNA methylation, a process in which a gene is kept inactive by addition of methyl groups to a DNA strand. The coats of agouti mice, for instance, can be either brown or a striking bright yellow, depending on the nutrient content of the maternal prenatal diet (Gilbert & Epel, 2009). The process by which the same genotypes can be expressed as strikingly different phenotypes as a result of environmental influences is known as epigenesis, and across species it can offer a powerful explanatory tool for the wide variations seen in genetically mediated conditions. In other words, it is not simply that children with a diet low in DHA/AA may have less brain tissue; it may also be true that their brains function less efficiently in certain contexts, and that the genes directing their future development may manifest their influence in very different ways. Many other human milk constituents may play a role in brain development, and they may work synergistically to optimize neurodevelopmental outcomes. The results of studies investigating LC-PUFA supplementation in formula-fed children lend support to this point of view. A meta-analysis by Simmer, Schulzke, and Patole (2008) reported that LC-PUFAsupplemented formula was not associated with neurodevelopmental benefit, indicating that DHA is unlikely to be the sole explanation for the differences observed between breastfed and formula-fed children. A related Cochrane review asserted that providing supplemental LC-PUFA to breastfeeding mothers did not result in improved cognition or language skills for their children (Delgado-Noguera, Calvache, & Bonfill Cosp, 2010). Clearly, then, it is prudent to consider other possible reasons why human milk might be associated with neurodevelopmental differences. One example of a nutrient that may play a role in neurodevelopment is choline, a substance sometimes known as Vitamin B4. In a 2006 review, Zeisel stated that choline is important for neural tube closure in human fetuses; animal research indicates that choline supplementation can have long-lasting effects on memory. Human milk is rich in choline. In infant formula, on the other hand, although the Food and Drug Administration (FDA) has established a minimum for choline content (FDA Infant Formula Nutrient Requirements, 2014), levels of this nutrient vary from one formulation to another (Zeisel & da Costa, 2009). In addition to choline, other human milk constituents have been identified as potentially important in neurodevelopment. In a 2010 investigation of brain volume in 50 adolescents, Isaacs and colleagues reported significant correlations between the percentage of breastmilk in the diet of infant boys, total brain volume, white matter volume, and verbal IQ. In other words, boys who received more human milk in infancy grew into adolescents with larger brains, more white matter, and stronger performance on tests of verbal IQ. (No significant effects were observed among the girls in the sample.) The
Please cite this article in press as: Mahurin-Smith, J. Breastfeeding and language outcomes: A review of the literature. Journal of Communication Disorders (2015), http://dx.doi.org/10.1016/j.jcomdis.2015.04.002
G Model
JCD-5703; No. of Pages 12 4
J. Mahurin-Smith / Journal of Communication Disorders xxx (2015) xxx–xxx
authors propose that multiple human milk constituents could promote brain growth, including nutrients such as cholesterol (which is absent from infant formula but potentially important for production of glial cells and myelin), hormones such as thyroxin, and additional elements such as nerve growth factor. Other compounds of interest could include carnitine, which plays a role in LC-PUFA metabolism; lactose, which is important for generating compounds called galactolipids thought to be crucial to CNS development (Lawrence & Lawrence, 2005); cytokines such as transforming growth factor-beta, which may influence regulation of inflammation (Oddy [13_TD$IF]& Rosales, 2010); and sialic acid, which may influence memory formation (Wang, 2009). The large array of substances found in human milk makes it difficult to investigate the differential effects of individual constituents. Furthermore, regulatory standards allow manufacturers some latitude in determining the levels of selected constituents; lactose, for instance, is absent from soy formula but abundant in formula based on cow’s milk. In addition to directly influencing brain structure and function through its nutrient profile, breastfeeding may exert more indirect effects as well. It is well established that early diet plays a role in long-term immune function (Wilson et al., 1998); it is less widely known that early diet affects the body’s ability to regulate inflammation (McDade, 2012). Shorter breastfeeding duration has been identified as a predictor for elevated levels of C-reactive protein (CRP), a widely recognized marker for inflammation (McDade et al., 2014). Wa¨rnberg, Gomez-Martinez, Romeo, Dı´az, and Marcos (2009) report that CRP has directly neurotoxic effects. Exploration of the relationship between inflammation and neurodevelopment is in its infancy, but dysregulation of immune function and inflammation has been reported in children with autism (Ashwood, Wils, & van de Water, 2006). Breastfeeding affects mothers as well as their infants, altering the maternal hormonal milieu (Lawrence & Lawrence, 2005), and these influences have the potential to inform speech-language outcomes as well. With every feeding, or 10–12 times per day for a fully breastfeeding or exclusively pumping mother, the pituitary repeatedly secretes prolactin and oxytocin to elicit milk production and ejection. Much of the available data on lactation and the maternal brain comes from animal studies. Although caution is essential in extrapolating from rodent dams to human mothers, rodent studies may have ramifications for human studies of breastfeeding outcomes. It has been known since the 1980s that oxytocin, the hormone that drives milk ejection, induces changes in maternal behavior in rats (see Pedersen, Ascher, Monroe, & Prange, 1982). More recent investigations have indicated that motherhood is associated with significant cognitive changes across many rodent studies: learning and memory are improved and related changes in brain function and/or morphology have been reported. Oxytocin has been suggested as a mediator of these changes (Darnaude´ry et al., 2007; Tomizawa et al., 2003), which may be long-lasting (Love et al., 2005). Human studies lend some support to the idea that lactation influences mothers’ brains and behavior, potentially shaping (1) maternal affect, (2) maternal brain structure and activation, and (3) mother–child interactions. On the topic of maternal affect, Mezzacappa and Katkin (2002) reported that breastfeeding can buffer maternal stress levels and negative maternal mood, even with control for demographic confounds. Regarding the maternal brain, anatomical changes have also been observed in the brains of lactating mothers, with postpartum gray matter growth in multiple regions including the prefrontal cortex (Kim et al., 2010). Kim et al. (2011) found that lactating mothers showed greater brain activation than formulafeeding mothers in the first month postpartum. In both groups, greater brain activation in the regions of interest was associated with improved maternal sensitivity when babies were 3–4 months old. This finding, together with reports that oxytocin may influence caregiver–infant relations (Feldman, 2007; Feldman, Weller, Zagoory-Sharon, & Levine, 2007; Gordon, Zagoory-Sharon, Leckman, & Feldman, 2010), suggests that breastfeeding might have long-lasting effects on mother–child interactions (see further discussion of this issue in Strathearn, Mamun, Najman, & O’Callaghan, 2009). Longitudinal studies of mother–infant dyads have reported positive correlations between maternal sensitivity during the infant period and later language development (Baumwell, Tamis-LeMonda, & Bornstein, 1997), with stronger effects observed for children at higher risk of developmental delay (Landry, Smith, Miller-Loncar, & Swank, 1997). While these relationships between lactation and maternal brain/behavior changes require further investigation, they may offer another lens through which to view the potential effects of infant feeding decisions. The next three sections of this paper will consider the existing literature on infant feeding as a predictor for language outcomes, first in population-based studies emphasizing typical development, and then in studies that focus on specific forms of communicative impairment, with a brief review of two well-publicized studies that reported null findings based on a single large dataset. Given the focus of this paper, no distinction is made between direct breastfeeding and feeding expressed milk. In both cases, the mother lactates and the child receives human milk. 4. The existing literature on typical development Very little of the research literature in communication sciences and disorders has evaluated infant diet as a possible environmental variable influencing speech-language development. There are, however, a number of studies from pediatrics, epidemiology, and related disciplines that have investigated this relationship. The question of whether breastfeeding might influence speech-language outcomes was first raised by Broad and colleagues in a series of 3 studies (Broad, 1972, 1975; Broad & Duganzich, 1983). All three studies reported that formula-fed children, particularly boys, were more likely to experience slower communicative development even with covariate control. These findings are mentioned here primarily for their historical interest, as recent advances in infant formula limit the extent to which extrapolation from Broad’s research is advisable.
Please cite this article in press as: Mahurin-Smith, J. Breastfeeding and language outcomes: A review of the literature. Journal of Communication Disorders (2015), http://dx.doi.org/10.1016/j.jcomdis.2015.04.002
G Model
JCD-5703; No. of Pages 12 J. Mahurin-Smith / Journal of Communication Disorders xxx (2015) xxx–xxx
5
More recent studies span the range of ages from infancy to adolescence. In a 1999 study of 1656 8-month-old Danish infants, Vestergaard and colleagues reported that 73.4% of the infants exclusively breastfed for 6 months were engaging in variegated babbling, versus 34.0% of the formula-fed infants. The overall odds ratio for emergence of variegated babbling was 1.5 (95% CI = 1.3–1.8). The relationship persisted with covariate control for factors including social class, maternal education, prenatal smoking, birthweight, gestational age, and number of prior illnesses. The authors reported that they purposefully chose to investigate differences in infancy, to reduce the period of time in which potential confounding variables such as maternal education could exert an influence. In a related study of 2302 Danish infants, Obel, Henriksen, Hedegaard, Secher, & Østergaard (1998) investigated the relationship between babbling in babyhood and prenatal smoking among mothers. They reported a dose-response relationship between prenatal exposure to cigarette smoke and decreased odds that babbling would emerge by 8 months, with an odds ratio (OR) of 2.0 (95% CI = 1.1–3.6) for maternal use of 10 cigarettes per day. They reported, however, that breastfeeding appeared to confer a measure of protection against the effects of prenatal cigarette smoke exposure on prelinguistic development. The infants who did not receive at least 4 months of exclusive breastfeeding were even less likely to be babbling at 8 months, OR = 2.7 (CI = 1.3–5.8). These two studies of pre-linguistic development appear to be unique in the breastfeeding literature; many authors, in contrast, have looked at neurodevelopment in older children through IQ scores. Since many cognitive assessment tools rely heavily on verbal skills, these studies can be very relevant to discussions of language ability. These findings, which generally show a modest but statistically robust difference in favor of breastfed children, have been summarized in other sources (see Anderson et al., 1999; Drane & Logemann, 2000). The remainder of this section will emphasize studies that have assessed language development with instruments developed by and for speech-language pathologists, with one exception at the end for an important 2008 study that used novel methodology to measure differences in verbal IQ. Thorpe, Rutter, and Greenwood (2003) studied 96 twin pairs in an attempt to find factors that predicted at 20 months how a child would perform on measures of language skill at 36 months. The twins were contrasted with 98 pairs of singleton siblings, because twins are known to develop language skills more slowly than singletons. The investigators collected data on breastfeeding duration, but not breastfeeding exclusivity. To evaluate language development, the investigators used the Preschool Language Scale–Third Edition (PLS-3; Zimmerman, Steiner, & Pond, 1992). The language skills of breastfed singletons were significantly stronger at both 20 and 36 months than those of the formula-fed singletons, but the effect ceased to be significant when the investigators controlled for maternal vocabulary. No association between breastfeeding and language was observed in the twin pairs. Gibson-Davis and Brooks-Gunn (2006) studied the children of 1645 American-born mothers, a subset of the Fragile Families and Child Wellbeing Survey cohort, to examine the impact of breastfeeding on the language skills of 3-year-old children. Raw PPVT-R scores were 6.2 points lower among bottle-fed children, a significant finding that remained so after adjustment for demographic factors, home environment, and health-related variables. When maternal PPVT-R scores were factored in, however, the difference was only marginally significant (p = .06). The authors described an interaction between breastfeeding and education as predictor variables, with a clearer effect of breastfeeding observed among more educated mothers. Gibson-Davis and Brooks-Gunn point out that their findings cannot be extrapolated to the general population, as the Fragile Families cohort includes a large proportion of disadvantaged children and differences are evident in the descriptive statistics for both the independent and dependent variables (IV and DV). Specifically, breastfeeding initiation and duration rates in this sample (the IV) were much lower than those in the population as a whole, leading the authors to use a bivariate classification for breastfeeding: breastfed 9 months faced an increased risk of parental concern about their language skills, need for SLP services, and scores >1 SD below the mean on the adapted PPVT-3. For children diagnosed with specific language impairment (SLI), breastfeeding appears to be a statistically robust predictor variable. Tomblin, Smith, & Zhang (1997) considered prenatal and perinatal risk factors for SLI in a case-control study with 1102 children, 177 of whom had an SLI diagnosis. These researchers evaluated children during kindergarten and interviewed their parents to assess factors potentially associated with SLI. Breastfeeding incidence and duration were lower
Please cite this article in press as: Mahurin-Smith, J. Breastfeeding and language outcomes: A review of the literature. Journal of Communication Disorders (2015), http://dx.doi.org/10.1016/j.jcomdis.2015.04.002
G Model
JCD-5703; No. of Pages 12 J. Mahurin-Smith / Journal of Communication Disorders xxx (2015) xxx–xxx
7
among children with SLI, even after differences in maternal education were accounted for. Children breastfed for less than nine months faced a significant increase in their risk of SLI, with an odds ratio of 2.5. In an unpublished dissertation, Drane (2003) re-analyzed the data from Tomblin et al.’s 1997 study while employing improved covariate control. She found that the odds ratio for SLI in formula-fed children declined only slightly, from 2.5 to approximately 2.4. Drane concluded that breastfeeding does appear to protect against SLI. Several studies have investigated the hypothesis that breastfeeding might protect against autism. Although they fall outside the scope of the present review, which is focused on language outcomes, interested readers are referred to a 1989 paper by Tanoue and Oda, replicated in 2006 by Schultz et al. Two other recent papers (Al-Farsi et al., 2012; Dodds et al., 2011) have reported on this relationship as well. In addition to effects on language and interaction, investigators have also described significant relationships between infant feeding and protection against speech impairment. These papers are not reviewed in full here, but interested readers are referred to Mahurin-Smith and Ambrose (2013), Barbosa et al. (2009), and Fox, Dodd, and Howard (2002). 6. Human neurodevelopment and infant diet: highly publicized null findings While the studies described in the preceding sections have generally reported statistically significant positive relationships between breastfeeding duration and neurodevelopmental outcomes, results in the literature on cognition have been more mixed and contentious. Two large studies from the past decade (Colen & Ramey, 2014; Der, Batty, & Deary, 2006) claimed that the purported influence of breastfeeding on cognition was in fact attributable to confounding variables, particularly maternal IQ. Because of the many similarities between measures of verbal IQ and measures of language development/impairment, the methodological issues raised by these studies are germane to the present review. These results have public health and policy implications, since it would be unhelpful to instruct mothers to breastfeed in hopes of improving neurodevelopmental outcomes if there is not in fact a relationship between the predictor and the outcome. For this reason the following section will briefly discuss the issues raised by Der et al. (2006) and Colen and Ramey (2014), both of which utilized data from the National Longitudinal Survey of Youth (NLSY) to evaluate the relationship between breastfeeding and cognition. The NLSY includes information on both maternal and child cognition, as well as information on feeding history for multiple siblings within participating families, and thus could be viewed as a natural way to evaluate the relationships among these variables in a large cohort. In their 2006 paper, Der and colleagues found substantial differences in maternal IQ for mothers who breastfed their children and children who formula-fed their children. When the researchers controlled for this difference, breastfeeding ceased to be a significant predictor of cognitive outcomes for their children. This study, along with the 2014 Colen and Ramey investigation, also considered the question of discordant sibling outcomes. If one sibling is breastfed and another is bottlefed, are there differences in IQ between the two children? Both studies reported that no significant differences were observed between discordant sibling pairs, and concluded from this finding that the importance of breastfeeding in long-term outcomes has been overstated. These studies raise three important issues. First is the need for definition of terms and for recognition of the impact of exclusivity and duration. No definition of breastfeeding is given in either of the two papers discussed in this section. If a mother breastfeeds her child once and decides to formula-feed thereafter, did she breastfeed or bottle-feed? As discussed previously, more consistent effects have been described in studies that accounted for breastfeeding exclusivity. A related question is the importance of breastfeeding duration. Some authors have reported a threshold effect with 3 months of breastfeeding (Barone et al., 2006; Dee et al., 2007), and it is thus noteworthy that the median duration of breastfeeding in the NLYS cohort was 3 months. If a study of acetaminophen users announced that the medication had proven ineffective for headache treatment, it would make a difference in the generalizability of the findings if half of the participants had only taken a fraction of a pill. Similarly, the conclusions that can be drawn from a sample in which half of the participants received less than the hypothesized threshold ‘‘dose’’ of human milk are necessarily limited. The second issue, mentioned above, is the issue of appropriate covariate control. It is well known that feeding decisions are influenced by extended family norms. Mothers who were themselves breastfed in infancy are significantly more likely to breastfeed their own children as adults (Meyerink & Marquis, 2002). If the 2008 PROBIT report from Kramer et al. is accurate, and breastfeeding is indeed associated with a boost of 0.5 SD in IQ, much of the 0.6 SD difference between groups in the Der, Batty, and Deary study would be explained. In a similar way, if the maternal capacity for learning and remembering is influenced by lactation, as discussed in the mechanism section, the question of how to control for differences in maternal ability becomes particularly thorny. While some form of covariate control is important, it may be imprudent to treat maternal IQ as a simple predictor to be partialed out. Further PROBIT reports or other randomized controlled trials may shed additional light on this question; observational studies from parts of the world in which breastfeeding is not associated with socioeconomic advantage (e.g., Daniels & Adair, 2005) may also be helpful in resolving these unanswered questions. A third issue raised by these two studies is the complexity of sibling comparisons. Although Colen and Ramey described discordant siblings as ‘‘a natural experiment,’’ the reality is that older and younger siblings differ from each other in both predictable and unpredictable ways. Comparisons of discordant sibling data must consider the context of maternal feeding decisions, as well as the constrained variability in outcomes inherent in a comparison of children who share, on average, 50% of their genetic material. Sibling comparisons may contribute something to the discussion of neurodevelopmental outcomes, but it would be premature to extrapolate too much from studies such as these.
Please cite this article in press as: Mahurin-Smith, J. Breastfeeding and language outcomes: A review of the literature. Journal of Communication Disorders (2015), http://dx.doi.org/10.1016/j.jcomdis.2015.04.002
G Model
JCD-5703; No. of Pages 12 8
J. Mahurin-Smith / Journal of Communication Disorders xxx (2015) xxx–xxx
7. Discussion 7.1. Review of findings The key points of this paper can be summarized as follows: first, it is reasonable to consider the influence of diet on language outcomes, as both human and animal studies have supported the idea of neurodevelopmental differences related to breastfeeding. Second, large population-based studies that consider language outcomes in relation to breastfeeding exclusivity have generally reported modest but statistically significant differences in favor of breastfed children. Third, studies of language impairment have consistently reported a protective effect of breastfeeding, with larger effect sizes noted in these populations. Finally, while null findings make occasional appearances in the headlines, caution is called for in interpreting studies of populations where breastfeeding exclusivity and duration are low and/or poorly documented. Across all of the studies described here, there is a trend for stronger associations to be observed in studies with a wider range of breastfeeding duration and a higher degree of breastfeeding exclusivity. Gibson-Davis and Brooks-Gunn (2006), with their
JCD-5703; No. of Pages 12 Journal of Communication Disorders xxx (2015) xxx–xxx
Contents lists available at ScienceDirect
Journal of Communication Disorders
Review
Breastfeeding and language outcomes: A review of the literature [9_TD$IF]J. Mahurin[10_TD$IF]-Smith * Illinois State University, Normal, IL 61790, United States
A R T I C L E I N F O
A B S T R A C T
Article history: Received 18 September 2014 Received in revised form 10 April 2015 Accepted 19 April 2015 Available online xxx
Many researchers have investigated the potential impact of breastfeeding in infancy on a child’s subsequent development, but only a small subset of these studies considers language development and impairment. This paper reviews that literature, discussing postnatal neurodevelopment, potential mechanisms for dietary influences on communication outcomes, studies of typically developing children, and studies of children with communication concerns. For population based studies of language development, a modest but statistically robust relationship is seen across large samples that account for breastfeeding exclusivity. A similar protective relationship is seen in studies that evaluate the relationship between breastfeeding and language disorders; effect sizes are typically larger in these papers. Implications for researchers and service providers are reviewed. Learning outcomes: Readers will be able to describe possible mechanisms by which early diet might influence neurodevelopment. They will be able to describe the relationships observed between diet in infancy and language outcomes in large population-based studies, as well as the trends observed in studies of the relationship between infant diet and communication impairment. ß 2015 Elsevier Inc. All rights reserved.
Keywords: Feeding Breastfeeding Child language Language delay Language disorder Neurodevelopment
Contents 1. 2. 3. 4. 5. 6. 7.
8.
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Postnatal neurodevelopment. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Why would diet influence language development? Potential mechanisms . . . . . . . . . . The existing literature on typical development . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . The existing literature on communication disorders . . . . . . . . . . . . . . . . . . . . . . . . . . . . Human neurodevelopment and infant diet: highly publicized null findings . . . . . . . . . Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7.1. Review of findings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7.2. Assessment of the existing research . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7.3. Implications for professionals in communication sciences and disorders (CSD) . Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
000 000 000 000 000 000 000 000 000 000 000 000 000
* Tel.: +1 309 438 5308. E-mail address: [email protected] http://dx.doi.org/10.1016/j.jcomdis.2015.04.002 0021-9924/ß 2015 Elsevier Inc. All rights reserved.
Please cite this article in press as: Mahurin-Smith, J. Breastfeeding and language outcomes: A review of the literature. Journal of Communication Disorders (2015), http://dx.doi.org/10.1016/j.jcomdis.2015.04.002
G Model
JCD-5703; No. of Pages 12 2
J. Mahurin-Smith / Journal of Communication Disorders xxx (2015) xxx–xxx
1. Introduction Imagine that two communities in two different parts of the world are building a network of roads. As a cost-saving measure, the planning teams opt to use lower-grade building materials for half of each community. One community’s climate is mild: temperatures tend to stay in the 40–70 8F range. The second community experiences much more temperature variability: winter temperatures might plummet to 20 8F or cycle rapidly from freeze to thaw and back again, while summer temperatures might hover around 100 8F. Drivers who have navigated pothole-ridden streets in spring might guess that the roads built with lower-grade materials could show signs of wear more rapidly in Community #2. When the strains on the materials are greater, the breakdowns in the network may be apparent more rapidly. Drivers in Community #1 might notice only minor signs of wear on the lowergrade sections of their roadways, while drivers in Community #2 might find a conspicuous difference. The neurological system can be envisioned as a vast network of roads, undergoing large-scale construction projects during pregnancy, infancy, and toddlerhood. The ‘‘road-building’’ materials available to children can vary with their early diet, leading to differences in brain structure (Farquharson, Jamieson, Logan, Cockburn, & Ainslie Patrick, 1992; Farquharson et al., 1995; Jamieson et al., 1999; Makrides, Neumann, Byard, Simmer, & Gibson, 1994). These differences in brain structure may be associated with differences in brain function (Greiner, Moriguchi, Hutton, Slotnick, & Salem, 1999; Salem et al., 2000). For children whose network of genetic and other environmental effects does not predispose them to language difficulties – similar to the roads in Community #1 – diet in infancy may be linked to modest differences in later life (Anderson, Johnstone, & Remley, 1999). For children whose genotype renders them more vulnerable to language impairment, like the roads in Community #2, early diet may make a more significant difference as they mature (Schultz et al., 2006). Breastfeeding is widely recommended for infants, and the differences in morbidity between breastfed and formula-fed children are well-documented (American Academy of Pediatrics, 2012; World Health Organization, 2003). Less familiar, perhaps, are the differences in neurodevelopment that have also been observed in studies of breastfed versus formula-fed children. Although a subset of this literature looks explicitly at the relationship between feeding mode in infancy and subsequent speech-language development, these investigations are seldom referenced in communication sciences and disorders (CSD) journals (Rogers et al., 2015). This paper will first describe the process of postnatal brain development and the mechanisms through which diet might influence neurodevelopment, followed by a review of the existing literature on associations between breastfeeding and subsequent language development and impairment. It will conclude with a review of potential implications arising from these findings. 2. Postnatal neurodevelopment Lawrence and Lawrence (2005) report that the average newborn will arrive with a brain weighing approximately 350 g; a year later, that brain will have grown to approximately 1100 g. By age 3 the brain will have quadrupled relative to its size at birth; the steepness of this early trajectory is particularly clear when one considers that the brain will require a further 15 years to approach the quintuple mark (Dekaban & Sadowsky, 1978). Neurons proliferate in regions including the dentate gyrus of the hippocampus, a region associated with memory; during the first 2 weeks of life this process is especially vulnerable to environmental influences such as the presence or absence of growth factors (Watson, DeSesso, Hurtt, & Cappon, 2006). In addition to the rapid growth of the brain itself, other types of neural tissue proliferate during the early years of life. The most rapid period of postnatal myelination occurs during the first 2 years of life; the major fiber tracts are clearly visible in 3-year-old children (Matsuzawa et al., 2001). During the first 18 months of life neurons must migrate to other parts of the brain, including the prefrontal cortex (Sanai et al., 2011). This is also a period of dramatic growth in synapse formation, so much so that at the point of peak synaptic density there are 55 synapses per 100 mm3 in the human brain (Watson et al., 2006). The decline in synaptic density observed between toddlerhood and adulthood is the result of a form of programmed cell death known as pruning. It is common to think of brain cell death as undesirable, but pruning is a critical part of normal neurodevelopment, distinct from other causes of cell death such as injury or toxin exposure (Watson et al., 2006). Inadequate pruning has been posited as a factor in autism (Hill & Frith, 2003), highlighting the potential importance of this neurodevelopmental process. During infancy the brain is also building its own protective mechanisms, a process that takes time to complete. In the first 6 months of life the blood-brain barrier is more permeable than it will be in later life, increasing the infant’s vulnerability to neurotoxins (e.g., methylmercury) with the potential to affect future cognitive skills (see Watson et al., 2006; Dzwilewski et al., in this issue). To sum up, then, infants not only must build brain cells, they must also establish connections between them, re-sculpt brain architecture, and create a selectively permeable wall between the brain and the outside world. Their diet provides the raw materials available for these tasks. 3. Why would diet influence language development? Potential mechanisms The idea that breastfeeding or formula-feeding could influence language development immediately raises the question of mechanism: why would children’s diets in infancy exert any long-term influence over their speech-language skills? Four possibilities will be reviewed in the section that follows. The explanation most often proposed is that differences in the fatty
Please cite this article in press as: Mahurin-Smith, J. Breastfeeding and language outcomes: A review of the literature. Journal of Communication Disorders (2015), http://dx.doi.org/10.1016/j.jcomdis.2015.04.002
G Model
JCD-5703; No. of Pages 12 J. Mahurin-Smith / Journal of Communication Disorders xxx (2015) xxx–xxx
3
acid profiles of human milk and infant formula might make a difference in subsequent brain structure and function; a closely related possibility is that the fatty acid content of cell membranes could influence gene expression within those cells. The second idea is that human milk might contain a variety of constituents that promote optimal neurodevelopment, perhaps with synergistic effects seen across multiple nutrients. A third consideration is that the well-documented relationship between breastfeeding and immune system function/regulation of inflammation might exert a moderating influence on learning and memory. A fourth potential mechanism involves the impact of lactation on mothers, who in turn shape their children’s learning and language. These mechanisms are not mutually exclusive, and the reader is reminded that a single causal mechanism is unlikely to provide an adequate explanation for a phenomenon as complex as neurodevelopment. Analysis of the brain tissue acquired during infancy indicates that more than half of its solid weight is lipid (Lawrence & Lawrence, 2005), and that the fatty acid profile of an infant’s diet plays a significant role in the composition of brain tissue (Farquharson et al., 1992, 1995; Jamieson et al., 1999; Makrides et al., 1994). Of particular importance are two long-chain polyunsaturated fatty acids: an omega-3 (or n-3) fatty acid known as docosahexaenoic acid (DHA) and an omega-6 (n-6) fatty acid called arachidonic acid (AA). Across mammal species, DHA is the most abundant fatty acid found in the brain, but the amounts of DHA in brain tissue are directly related to an individual’s diet (Innis, 2007). Not all of the infant formulas for sale in the US contain DHA and AA; those that do are consistently more expensive than formulations with supplemental LC-PUFA. For this reason, formula-fed infants in the US may lack a dietary source of DHA and AA. Much of the evidence supporting the relationships among dietary fatty acid intake, brain composition, and learning/ behavioral outcomes comes from animal studies, in which diet can be manipulated to a degree that would be gravely unethical in studies of human infants. It has long been documented that rat pups fed a diet free of n-3 fatty acids build brain tissue with much less DHA and a corresponding increase in n-6 fatty acids (Bourre et al., 1984). These differences in n-3 intake are associated with poorer learning and memory impairment in adult rats (Greiner et al., 1999), as well as worse scores on measures of depression and aggression (DeMar et al., 2006). Studies of human children also suggest that the fatty acid content of the diet influences the structure of neural tissue, both gray matter and white matter. Kafouri et al. (2012) investigated cortical thickness in 571 adolescents and found that increased duration of exclusive breastfeeding was strongly correlated with both increased thickness of the parietal cortex and with higher cognitive scores; the authors posited that the DHA in human milk may explain these findings. Another recent study (Deoni et al., [12_TD$IF]2013) found slower myelination in formula-fed children as compared to their breastfed counterparts. These authors state, ‘‘While the exact mechanism(s) that underlie these. . .differences remain unclear, our results lend support to the hypothesis that the docosahexaenoic and arachidonic acids present in breast milk promote preferential neural growth and white matter development’’ (Deoni et al., 2013, p. 83). Gross quantitative measures of neural tissue offer only a rough index of brain maturation. At a microscopic level, LC-PUFA intake influences the growth of dendritic spines and the creation of synaptic membranes, thus shaping neurotransmission and cell-to-cell signaling (Nyaradi, Li, Hickling, Foster, & Oddy, 2013). For scholars of genetically mediated phenomena such as speech and language, it is perhaps even more important to know that the fatty acid content of cell membranes can shape gene transcription and expression (Innis, 2007). Animal studies have indicated that dietary intake is well-documented as an agent of DNA methylation, a process in which a gene is kept inactive by addition of methyl groups to a DNA strand. The coats of agouti mice, for instance, can be either brown or a striking bright yellow, depending on the nutrient content of the maternal prenatal diet (Gilbert & Epel, 2009). The process by which the same genotypes can be expressed as strikingly different phenotypes as a result of environmental influences is known as epigenesis, and across species it can offer a powerful explanatory tool for the wide variations seen in genetically mediated conditions. In other words, it is not simply that children with a diet low in DHA/AA may have less brain tissue; it may also be true that their brains function less efficiently in certain contexts, and that the genes directing their future development may manifest their influence in very different ways. Many other human milk constituents may play a role in brain development, and they may work synergistically to optimize neurodevelopmental outcomes. The results of studies investigating LC-PUFA supplementation in formula-fed children lend support to this point of view. A meta-analysis by Simmer, Schulzke, and Patole (2008) reported that LC-PUFAsupplemented formula was not associated with neurodevelopmental benefit, indicating that DHA is unlikely to be the sole explanation for the differences observed between breastfed and formula-fed children. A related Cochrane review asserted that providing supplemental LC-PUFA to breastfeeding mothers did not result in improved cognition or language skills for their children (Delgado-Noguera, Calvache, & Bonfill Cosp, 2010). Clearly, then, it is prudent to consider other possible reasons why human milk might be associated with neurodevelopmental differences. One example of a nutrient that may play a role in neurodevelopment is choline, a substance sometimes known as Vitamin B4. In a 2006 review, Zeisel stated that choline is important for neural tube closure in human fetuses; animal research indicates that choline supplementation can have long-lasting effects on memory. Human milk is rich in choline. In infant formula, on the other hand, although the Food and Drug Administration (FDA) has established a minimum for choline content (FDA Infant Formula Nutrient Requirements, 2014), levels of this nutrient vary from one formulation to another (Zeisel & da Costa, 2009). In addition to choline, other human milk constituents have been identified as potentially important in neurodevelopment. In a 2010 investigation of brain volume in 50 adolescents, Isaacs and colleagues reported significant correlations between the percentage of breastmilk in the diet of infant boys, total brain volume, white matter volume, and verbal IQ. In other words, boys who received more human milk in infancy grew into adolescents with larger brains, more white matter, and stronger performance on tests of verbal IQ. (No significant effects were observed among the girls in the sample.) The
Please cite this article in press as: Mahurin-Smith, J. Breastfeeding and language outcomes: A review of the literature. Journal of Communication Disorders (2015), http://dx.doi.org/10.1016/j.jcomdis.2015.04.002
G Model
JCD-5703; No. of Pages 12 4
J. Mahurin-Smith / Journal of Communication Disorders xxx (2015) xxx–xxx
authors propose that multiple human milk constituents could promote brain growth, including nutrients such as cholesterol (which is absent from infant formula but potentially important for production of glial cells and myelin), hormones such as thyroxin, and additional elements such as nerve growth factor. Other compounds of interest could include carnitine, which plays a role in LC-PUFA metabolism; lactose, which is important for generating compounds called galactolipids thought to be crucial to CNS development (Lawrence & Lawrence, 2005); cytokines such as transforming growth factor-beta, which may influence regulation of inflammation (Oddy [13_TD$IF]& Rosales, 2010); and sialic acid, which may influence memory formation (Wang, 2009). The large array of substances found in human milk makes it difficult to investigate the differential effects of individual constituents. Furthermore, regulatory standards allow manufacturers some latitude in determining the levels of selected constituents; lactose, for instance, is absent from soy formula but abundant in formula based on cow’s milk. In addition to directly influencing brain structure and function through its nutrient profile, breastfeeding may exert more indirect effects as well. It is well established that early diet plays a role in long-term immune function (Wilson et al., 1998); it is less widely known that early diet affects the body’s ability to regulate inflammation (McDade, 2012). Shorter breastfeeding duration has been identified as a predictor for elevated levels of C-reactive protein (CRP), a widely recognized marker for inflammation (McDade et al., 2014). Wa¨rnberg, Gomez-Martinez, Romeo, Dı´az, and Marcos (2009) report that CRP has directly neurotoxic effects. Exploration of the relationship between inflammation and neurodevelopment is in its infancy, but dysregulation of immune function and inflammation has been reported in children with autism (Ashwood, Wils, & van de Water, 2006). Breastfeeding affects mothers as well as their infants, altering the maternal hormonal milieu (Lawrence & Lawrence, 2005), and these influences have the potential to inform speech-language outcomes as well. With every feeding, or 10–12 times per day for a fully breastfeeding or exclusively pumping mother, the pituitary repeatedly secretes prolactin and oxytocin to elicit milk production and ejection. Much of the available data on lactation and the maternal brain comes from animal studies. Although caution is essential in extrapolating from rodent dams to human mothers, rodent studies may have ramifications for human studies of breastfeeding outcomes. It has been known since the 1980s that oxytocin, the hormone that drives milk ejection, induces changes in maternal behavior in rats (see Pedersen, Ascher, Monroe, & Prange, 1982). More recent investigations have indicated that motherhood is associated with significant cognitive changes across many rodent studies: learning and memory are improved and related changes in brain function and/or morphology have been reported. Oxytocin has been suggested as a mediator of these changes (Darnaude´ry et al., 2007; Tomizawa et al., 2003), which may be long-lasting (Love et al., 2005). Human studies lend some support to the idea that lactation influences mothers’ brains and behavior, potentially shaping (1) maternal affect, (2) maternal brain structure and activation, and (3) mother–child interactions. On the topic of maternal affect, Mezzacappa and Katkin (2002) reported that breastfeeding can buffer maternal stress levels and negative maternal mood, even with control for demographic confounds. Regarding the maternal brain, anatomical changes have also been observed in the brains of lactating mothers, with postpartum gray matter growth in multiple regions including the prefrontal cortex (Kim et al., 2010). Kim et al. (2011) found that lactating mothers showed greater brain activation than formulafeeding mothers in the first month postpartum. In both groups, greater brain activation in the regions of interest was associated with improved maternal sensitivity when babies were 3–4 months old. This finding, together with reports that oxytocin may influence caregiver–infant relations (Feldman, 2007; Feldman, Weller, Zagoory-Sharon, & Levine, 2007; Gordon, Zagoory-Sharon, Leckman, & Feldman, 2010), suggests that breastfeeding might have long-lasting effects on mother–child interactions (see further discussion of this issue in Strathearn, Mamun, Najman, & O’Callaghan, 2009). Longitudinal studies of mother–infant dyads have reported positive correlations between maternal sensitivity during the infant period and later language development (Baumwell, Tamis-LeMonda, & Bornstein, 1997), with stronger effects observed for children at higher risk of developmental delay (Landry, Smith, Miller-Loncar, & Swank, 1997). While these relationships between lactation and maternal brain/behavior changes require further investigation, they may offer another lens through which to view the potential effects of infant feeding decisions. The next three sections of this paper will consider the existing literature on infant feeding as a predictor for language outcomes, first in population-based studies emphasizing typical development, and then in studies that focus on specific forms of communicative impairment, with a brief review of two well-publicized studies that reported null findings based on a single large dataset. Given the focus of this paper, no distinction is made between direct breastfeeding and feeding expressed milk. In both cases, the mother lactates and the child receives human milk. 4. The existing literature on typical development Very little of the research literature in communication sciences and disorders has evaluated infant diet as a possible environmental variable influencing speech-language development. There are, however, a number of studies from pediatrics, epidemiology, and related disciplines that have investigated this relationship. The question of whether breastfeeding might influence speech-language outcomes was first raised by Broad and colleagues in a series of 3 studies (Broad, 1972, 1975; Broad & Duganzich, 1983). All three studies reported that formula-fed children, particularly boys, were more likely to experience slower communicative development even with covariate control. These findings are mentioned here primarily for their historical interest, as recent advances in infant formula limit the extent to which extrapolation from Broad’s research is advisable.
Please cite this article in press as: Mahurin-Smith, J. Breastfeeding and language outcomes: A review of the literature. Journal of Communication Disorders (2015), http://dx.doi.org/10.1016/j.jcomdis.2015.04.002
G Model
JCD-5703; No. of Pages 12 J. Mahurin-Smith / Journal of Communication Disorders xxx (2015) xxx–xxx
5
More recent studies span the range of ages from infancy to adolescence. In a 1999 study of 1656 8-month-old Danish infants, Vestergaard and colleagues reported that 73.4% of the infants exclusively breastfed for 6 months were engaging in variegated babbling, versus 34.0% of the formula-fed infants. The overall odds ratio for emergence of variegated babbling was 1.5 (95% CI = 1.3–1.8). The relationship persisted with covariate control for factors including social class, maternal education, prenatal smoking, birthweight, gestational age, and number of prior illnesses. The authors reported that they purposefully chose to investigate differences in infancy, to reduce the period of time in which potential confounding variables such as maternal education could exert an influence. In a related study of 2302 Danish infants, Obel, Henriksen, Hedegaard, Secher, & Østergaard (1998) investigated the relationship between babbling in babyhood and prenatal smoking among mothers. They reported a dose-response relationship between prenatal exposure to cigarette smoke and decreased odds that babbling would emerge by 8 months, with an odds ratio (OR) of 2.0 (95% CI = 1.1–3.6) for maternal use of 10 cigarettes per day. They reported, however, that breastfeeding appeared to confer a measure of protection against the effects of prenatal cigarette smoke exposure on prelinguistic development. The infants who did not receive at least 4 months of exclusive breastfeeding were even less likely to be babbling at 8 months, OR = 2.7 (CI = 1.3–5.8). These two studies of pre-linguistic development appear to be unique in the breastfeeding literature; many authors, in contrast, have looked at neurodevelopment in older children through IQ scores. Since many cognitive assessment tools rely heavily on verbal skills, these studies can be very relevant to discussions of language ability. These findings, which generally show a modest but statistically robust difference in favor of breastfed children, have been summarized in other sources (see Anderson et al., 1999; Drane & Logemann, 2000). The remainder of this section will emphasize studies that have assessed language development with instruments developed by and for speech-language pathologists, with one exception at the end for an important 2008 study that used novel methodology to measure differences in verbal IQ. Thorpe, Rutter, and Greenwood (2003) studied 96 twin pairs in an attempt to find factors that predicted at 20 months how a child would perform on measures of language skill at 36 months. The twins were contrasted with 98 pairs of singleton siblings, because twins are known to develop language skills more slowly than singletons. The investigators collected data on breastfeeding duration, but not breastfeeding exclusivity. To evaluate language development, the investigators used the Preschool Language Scale–Third Edition (PLS-3; Zimmerman, Steiner, & Pond, 1992). The language skills of breastfed singletons were significantly stronger at both 20 and 36 months than those of the formula-fed singletons, but the effect ceased to be significant when the investigators controlled for maternal vocabulary. No association between breastfeeding and language was observed in the twin pairs. Gibson-Davis and Brooks-Gunn (2006) studied the children of 1645 American-born mothers, a subset of the Fragile Families and Child Wellbeing Survey cohort, to examine the impact of breastfeeding on the language skills of 3-year-old children. Raw PPVT-R scores were 6.2 points lower among bottle-fed children, a significant finding that remained so after adjustment for demographic factors, home environment, and health-related variables. When maternal PPVT-R scores were factored in, however, the difference was only marginally significant (p = .06). The authors described an interaction between breastfeeding and education as predictor variables, with a clearer effect of breastfeeding observed among more educated mothers. Gibson-Davis and Brooks-Gunn point out that their findings cannot be extrapolated to the general population, as the Fragile Families cohort includes a large proportion of disadvantaged children and differences are evident in the descriptive statistics for both the independent and dependent variables (IV and DV). Specifically, breastfeeding initiation and duration rates in this sample (the IV) were much lower than those in the population as a whole, leading the authors to use a bivariate classification for breastfeeding: breastfed 9 months faced an increased risk of parental concern about their language skills, need for SLP services, and scores >1 SD below the mean on the adapted PPVT-3. For children diagnosed with specific language impairment (SLI), breastfeeding appears to be a statistically robust predictor variable. Tomblin, Smith, & Zhang (1997) considered prenatal and perinatal risk factors for SLI in a case-control study with 1102 children, 177 of whom had an SLI diagnosis. These researchers evaluated children during kindergarten and interviewed their parents to assess factors potentially associated with SLI. Breastfeeding incidence and duration were lower
Please cite this article in press as: Mahurin-Smith, J. Breastfeeding and language outcomes: A review of the literature. Journal of Communication Disorders (2015), http://dx.doi.org/10.1016/j.jcomdis.2015.04.002
G Model
JCD-5703; No. of Pages 12 J. Mahurin-Smith / Journal of Communication Disorders xxx (2015) xxx–xxx
7
among children with SLI, even after differences in maternal education were accounted for. Children breastfed for less than nine months faced a significant increase in their risk of SLI, with an odds ratio of 2.5. In an unpublished dissertation, Drane (2003) re-analyzed the data from Tomblin et al.’s 1997 study while employing improved covariate control. She found that the odds ratio for SLI in formula-fed children declined only slightly, from 2.5 to approximately 2.4. Drane concluded that breastfeeding does appear to protect against SLI. Several studies have investigated the hypothesis that breastfeeding might protect against autism. Although they fall outside the scope of the present review, which is focused on language outcomes, interested readers are referred to a 1989 paper by Tanoue and Oda, replicated in 2006 by Schultz et al. Two other recent papers (Al-Farsi et al., 2012; Dodds et al., 2011) have reported on this relationship as well. In addition to effects on language and interaction, investigators have also described significant relationships between infant feeding and protection against speech impairment. These papers are not reviewed in full here, but interested readers are referred to Mahurin-Smith and Ambrose (2013), Barbosa et al. (2009), and Fox, Dodd, and Howard (2002). 6. Human neurodevelopment and infant diet: highly publicized null findings While the studies described in the preceding sections have generally reported statistically significant positive relationships between breastfeeding duration and neurodevelopmental outcomes, results in the literature on cognition have been more mixed and contentious. Two large studies from the past decade (Colen & Ramey, 2014; Der, Batty, & Deary, 2006) claimed that the purported influence of breastfeeding on cognition was in fact attributable to confounding variables, particularly maternal IQ. Because of the many similarities between measures of verbal IQ and measures of language development/impairment, the methodological issues raised by these studies are germane to the present review. These results have public health and policy implications, since it would be unhelpful to instruct mothers to breastfeed in hopes of improving neurodevelopmental outcomes if there is not in fact a relationship between the predictor and the outcome. For this reason the following section will briefly discuss the issues raised by Der et al. (2006) and Colen and Ramey (2014), both of which utilized data from the National Longitudinal Survey of Youth (NLSY) to evaluate the relationship between breastfeeding and cognition. The NLSY includes information on both maternal and child cognition, as well as information on feeding history for multiple siblings within participating families, and thus could be viewed as a natural way to evaluate the relationships among these variables in a large cohort. In their 2006 paper, Der and colleagues found substantial differences in maternal IQ for mothers who breastfed their children and children who formula-fed their children. When the researchers controlled for this difference, breastfeeding ceased to be a significant predictor of cognitive outcomes for their children. This study, along with the 2014 Colen and Ramey investigation, also considered the question of discordant sibling outcomes. If one sibling is breastfed and another is bottlefed, are there differences in IQ between the two children? Both studies reported that no significant differences were observed between discordant sibling pairs, and concluded from this finding that the importance of breastfeeding in long-term outcomes has been overstated. These studies raise three important issues. First is the need for definition of terms and for recognition of the impact of exclusivity and duration. No definition of breastfeeding is given in either of the two papers discussed in this section. If a mother breastfeeds her child once and decides to formula-feed thereafter, did she breastfeed or bottle-feed? As discussed previously, more consistent effects have been described in studies that accounted for breastfeeding exclusivity. A related question is the importance of breastfeeding duration. Some authors have reported a threshold effect with 3 months of breastfeeding (Barone et al., 2006; Dee et al., 2007), and it is thus noteworthy that the median duration of breastfeeding in the NLYS cohort was 3 months. If a study of acetaminophen users announced that the medication had proven ineffective for headache treatment, it would make a difference in the generalizability of the findings if half of the participants had only taken a fraction of a pill. Similarly, the conclusions that can be drawn from a sample in which half of the participants received less than the hypothesized threshold ‘‘dose’’ of human milk are necessarily limited. The second issue, mentioned above, is the issue of appropriate covariate control. It is well known that feeding decisions are influenced by extended family norms. Mothers who were themselves breastfed in infancy are significantly more likely to breastfeed their own children as adults (Meyerink & Marquis, 2002). If the 2008 PROBIT report from Kramer et al. is accurate, and breastfeeding is indeed associated with a boost of 0.5 SD in IQ, much of the 0.6 SD difference between groups in the Der, Batty, and Deary study would be explained. In a similar way, if the maternal capacity for learning and remembering is influenced by lactation, as discussed in the mechanism section, the question of how to control for differences in maternal ability becomes particularly thorny. While some form of covariate control is important, it may be imprudent to treat maternal IQ as a simple predictor to be partialed out. Further PROBIT reports or other randomized controlled trials may shed additional light on this question; observational studies from parts of the world in which breastfeeding is not associated with socioeconomic advantage (e.g., Daniels & Adair, 2005) may also be helpful in resolving these unanswered questions. A third issue raised by these two studies is the complexity of sibling comparisons. Although Colen and Ramey described discordant siblings as ‘‘a natural experiment,’’ the reality is that older and younger siblings differ from each other in both predictable and unpredictable ways. Comparisons of discordant sibling data must consider the context of maternal feeding decisions, as well as the constrained variability in outcomes inherent in a comparison of children who share, on average, 50% of their genetic material. Sibling comparisons may contribute something to the discussion of neurodevelopmental outcomes, but it would be premature to extrapolate too much from studies such as these.
Please cite this article in press as: Mahurin-Smith, J. Breastfeeding and language outcomes: A review of the literature. Journal of Communication Disorders (2015), http://dx.doi.org/10.1016/j.jcomdis.2015.04.002
G Model
JCD-5703; No. of Pages 12 8
J. Mahurin-Smith / Journal of Communication Disorders xxx (2015) xxx–xxx
7. Discussion 7.1. Review of findings The key points of this paper can be summarized as follows: first, it is reasonable to consider the influence of diet on language outcomes, as both human and animal studies have supported the idea of neurodevelopmental differences related to breastfeeding. Second, large population-based studies that consider language outcomes in relation to breastfeeding exclusivity have generally reported modest but statistically significant differences in favor of breastfed children. Third, studies of language impairment have consistently reported a protective effect of breastfeeding, with larger effect sizes noted in these populations. Finally, while null findings make occasional appearances in the headlines, caution is called for in interpreting studies of populations where breastfeeding exclusivity and duration are low and/or poorly documented. Across all of the studies described here, there is a trend for stronger associations to be observed in studies with a wider range of breastfeeding duration and a higher degree of breastfeeding exclusivity. Gibson-Davis and Brooks-Gunn (2006), with their
