Updated: Aug 25
Dr Aimee Grant, Swansea University.
A recent review of research evidence suggested that “not breastfeeding is a risk factor for ASD” (ASD = Autistic Spectrum Disorder). Accordingly, the authors noted “the importance of breastfeeding in decreasing the risk of ASD among children”. As an Autistic breastfeeding researcher, I’m aware that breastmilk is pretty impressive stuff, but could it really prevent Autism as the researchers claim? No. Just like the MMR vaccine (which could not cause Autism), breastfeeding absolutely could not prevent Autism, and I’m going to tell you why.
First, I’m going to give you a very quick stats lesson, but don’t worry - I’ll make it less painful by talking about ice cream! When we do research and want to say that Thing A causes Thing B, we have to do a study where the researchers can control who gets to have Thing A and who does not. Not only does the researcher need to control it, but they also need to randomly assign who goes into each group, for example by flipping a coin – heads gets Thing A, tails does not. If researchers can’t randomly assign people to groups, the results may show a statistical relationship (also known as a correlation) between the two things, but it can’t be said that Thing A caused Thing B and so these results could be completely meaningless. For example, the more ice creams are consumed, the higher the murder rate but it would be completely wrong to say that eating ice cream causes murder (you can see more about the statistical dangers of ice cream here)! The mistake of believing that a statistical relationship between two things always means that Thing A is causing Thing B is the mistake made by the authors of the research. We don’t always know why the relationships exists.
Second, I’m going to show you that the relationship the researchers found between not breastfeeding and an Autism diagnosis is relatively weak. When doing statistics, the bigger your sample size, the more certainty you can have that your answer is correct for a general population of people. This study only had 3,270 individuals. For some diseases, this would be considered quite a high sample size, but when you consider that around 600,000 babies are born each year just in the UK and we’d expect 1-2% of the babies to be Autistic (6,000-12,000), and the fact that this study searched for all evidence internationally, that sample size is quite small. When researchers say that there is a statistical relationship between two things, we give something called a “confidence interval”; that is where we are sure the true answer lies between. With a smaller sample size you get a larger confidence interval. In this study the researchers said that babies who were not breastfed had 1.81 times the likelihood of being Autistic (so for every 1 breastfed Autistic child, there would be 1.81 formula fed Autistic children). But when we look at their confidence interval, this goes between 1.35 times and 2.27 times the likelihood. This isn’t bad science, but it shows that the evidence is limited.
Third, I’m not convinced that their measure of “breastfeeding” is useful. The definition of “ever breastfeeding” that is used is that a birthing parent has put the baby to the breast and tried to feed them at least once. This was once described to me by a UK midwife as “wafting the baby somewhere in the direction of the nipple”. So, consider the birthing parent is in hospital: they don’t have great breastfeeding support as the midwives are understaffed or not well trained, and they’ve tried a few times to breastfeed, but don’t think they’ve got the hang of it. In order to be discharged from the busy, loud postnatal ward so they can actually get some rest, they’re told they need to either breastfeed successfully or formula feed the baby, so the baby is given formula. Another birthing person may have managed to breastfeed in hospital and goes home, but then struggles and doesn’t have any breastfeeding support - after a few weeks they move on to formula. Birthing parent 3 has breastfed their previous baby, has a friend who’s a breastfeeding peer supporter and breastfeeds the child exclusively for 6 months, and then carries on until the infant is over two years old. All three of these babies would be counted as having been breastfed by the NHS, and you can see that there are vastly different lengths and intensities of breastfeeding for each of these children. What this means is that in the UK we’d expect over 80% of babies to be classed as breastfeeding at birth (wafting the nipple), and the UK has some of the lowest rates of breastfeeding in Europe so a group classed as “not breastfeeding” in other European countries would be really quite a small group.
In the review paper that is available online, there is no detail about how “ever breastfeeding” was measured; would wafting the nipple once have counted as breastfeeding? We also don’t know how many children were in each group (breastfeeding/not breastfeeding) in the wider population the studies took their samples from; were the not breastfeeding group as small as they would have been in the UK, or even smaller? Aside from the country they live in, we also do not know any demographic factors that were known about the birthing parents or babies. The authors of the review report that in five of the studies, babies (or their birthing parent) who didn’t breastfeed were matched with babies (or their birthing parent) who were similar to them in terms of demographics who did breastfeed. The reviewers, however, do not report the ways in which the pairs were matched (e.g.: was maternal age considered but being a single parent not?). Regardless of the number of ways in which people were paired, we know that a lot of factors make a difference to breastfeeding behaviour. We also know that there is a genetic link, with Autism running in families, and that Autistic parents can find breastfeeding challenging in ways that neurotypical birthing parents don’t report.
Lastly, I’m not sure how useful their definition of Autistic is. The CDC say that 1 in 44 children aged 8 are diagnosed as Autistic in their medical records. Getting access to an Autism diagnosis can be challenging, and girls are less likely to be diagnosed than boys. One of the studies in the review allowed for self-reporting of Autism status, but five used diagnoses based on the Diagnostic and Statistical Manual. The age of the children within the studies included in the review ranged from 2-18 years; Accordingly, it is likely that some Autistic children were not diagnosed when they were matched as pairs when they were young.
So, where does that leave us? It would be unethical to do a randomised trial where some mothers are told to breastfeed and others to formula feed, as we know that there are health benefits of breastfeeding for baby and parent. What this means is that we will never be able to say that not breastfeeding is a “risk” for Autism, even if there is an association. At the moment, the data that this conclusion is based on doesn’t feel that they’ve been robustly enough interrogated by the authors for me to feel confident that there is definitely an association. What we do know is that breastfeeding can be more challenging for Autistic parents, and that there is inadequate breastfeeding support in the UK for all families. I’d recommend energy is spent on finding ways to help those who want to breastfeed meet their breastfeeding goals. For Autistic parents, breastfeeding support may need to be tailored to account for sensory and interoceptive differences, and variation in pain experiences.