Source: BabyCenter USA

News outlets around the West are once again rushing out to share new research that suggests feeding infants solid foods as early as 3 months could improve infant sleep.  Before we even get to the actual research at hand, there are a couple things that are worth pointing out more generally about this research.  First and foremost, parents (and heck, all of Western society) need to realize that in the vast majority of cases there is nothing “wrong” with infants and their sleep and when there is something wrong (like sleep apnea, for example), introducing solids would be about as effective as talking to a banana to try and fix it.  Infants do not sleep as adults do and though we don’t know all the reasons why this is (though we know some like how sleeping too deeply increases the risk for SIDS), you can bet there’s a very good reason for it from an evolutionary or biological perspective.

Second, babies should not be having solids at 3 months.  Currently the recommendation is 6 months from the World Health Organization and their reasoning is pretty sound.  That said, I would be remiss if I didn’t mention that some cultures around the world have – and continue to – introduce very bland foods as young as 4 months.  Babies guts are not developed enough to handle solids at 3 months and certainly not developed to handle a lot of the crap that we give them.  The idea of what we feed them may be critically important as we look to the short- and long-term effects, but sadly we have little research on exactly how the processing and preserving of infant foods or the pesticide use on certain grains may be impacting their health.  What do we know?

  • The “benefits” of breastfeeding are greatest when done exclusively for 6 months
  • Babies don’t have the enzymes to digest certain foods – like starches – prior to the 4-6 month time (this is individual and thus each child is different)
  • Before 4 months, babies use their tongues to push food out of their mouths and lack coordination to swallow effectively and thus are at higher risk of choking
  • They are at an increased risk of obesity when fed solids too early
  • Eating solids early can actually result in malnutrition as they get full from the solid foods they can’t properly digest and thus intake less of the high-density (nutrition-wise) foods of breast milk or formula

So overall this seems like a pretty stupid idea to begin with, but here we are with a push to feed babies earlier – along with the associated risks above – in order to “help” their sleep which is actually probably biologically normal and healthy for them.  I suppose we should at least take a look at the research itself.

The Study

The study (full text linked here) was actually a secondary analysis of a larger study called the Enquiring About Tolerance study which was a population-based, randomized clinical trial conducted in England (on infants from England and Wales) on the introduction of allergenic foods and outcomes.  The current analyses were on 1303 infants who were exclusively breastfed.  The intervention group – called the early introduction group (EIG) – were told to introduce nonallergenic and then allergenic foods as early as 3 months of age while the control group – called the standard introduction group (SIG) – were given the advice to follow the current guidelines of 6 months of exclusive breastfeeding.  Infant sleep was assessed using the Brief Infant Sleep Questionnaire and was not measured using any objective measure.

The Findings

The first question we have to ask is adherence to protocol; that is, how many families actually did what they were told?  In this case, for those participants they had data for, the adherence in the SIG group was quite high at 92.1% but was much lower in the EIG group at 42.8%; however, it is worth noting that the protocol refers to the following for the SID group:

  • Exclusive breastfeeding for 3 months and breastfeeding up to 5 months or beyond
  • No consumption of allergenic foods before 5 months (this includes peanut, egg, sesame, fish, or wheat)
  • Less than 300mls/day of cow or goat’s milk or formula

For the EID group:

  • Exclusive breastfeeding for 3 months and breastfeeding up to 5 months or beyond
  • Consumption of at least 5 of the allergenic foods in at least 75% of the recommended amount (3g allergen protein/week) for at least 5 weeks between 3 and 6 months

In terms of the introduction of solid foods, the majority of families in the SIG group introduced solids before 6 months with 25% introducing them earlier than 21 weeks.  There was a significant difference in the average age of onset of solids (23.1 weeks versus 16. 2 weeks).

A quick, important aside about the statistical techniques…

The analyses for sleep-related outcomes in the primary paper were done using “intent-to-treat” (ITT) methods which means that individuals were included in their group regardless of how well they adhered to any protocol.  These are common when we have population-based randomization because they allow the randomization to remain.  If we focus on those who actually adhered to the protocol, then we may not have true randomization and may overestimate the effect of an intervention.  Given that randomization is key to many statistical procedures to reduce bias (it allows to us to ignore certain assumptions about our groups), ITT is often applied in clinical trials.

However, what this means is that the findings do not speak to the specific intervention protocol, but rather the potential effects of a treatment policy.  This means that we actually do not have data on the early introduction of solids here, but we have data on what some kind of intervention done early on feeding may result in.  This is key because everything else that follows has to be viewed in this lens.  It also could mean that any intervention protocol (like saying to only introduce potatoes) might have the same effect.  If they wanted to test the efficacy of the actual intervention protocol then they would need to look at those who completed the protocol (referred to as a per-protocol analysis which was done in the supplemental section which I will discuss below as well).

Of note, ITT is considered excellent in medical research because providing an intervention is relatively straightforward and one cannot choose to take an intervention without cause.  When used in social science, like with feeding and sleep, where parents have a choice to engage in an intervention, it’s less clear that this is as beneficial a statistical technique.  When parents read about a study like this, they may choose to introduce solids earlier, but this misses that the intervention itself was not appropriately analyzed because of the ITT method.

So back to those findings…

The researchers found that infants in the EIG group slept for an average of 7.3 minutes more per night than those in the SIG group.  When other factors were included in the model, the peak difference was 16.6 minutes at 6 months of age (notably when the other group was closer to introducing solids which often results in disrupted sleep).  There were also differences in reported night wakings (9.1% fewer in the EIG group) and parent-reported sleep problems (they were greater for both minor and major problems in the SIG group).

Now to that per-protocol analysis in the supplemental section…

The authors did do the per-protocol analysis, but of note this referred to the introduction of allergenic foods.  On the surface we have to ask ourselves if we believe that allergenic foods should improve sleep – what is the theory behind this?  We ask this because if we find a significant result, we have to think about the likelihood that it actually reflects chance or the fact that with such a large sample size, very small differences are significant.

In this case, there were significant effects on sleep based on the graphs provided (no analyses actually given).  Basically, the effects were far greater than in the ITT analyses and suggested that it’s not the introduction of solids, but the introduction of allergenic solids that may improve sleep.  However, even these average numbers were questionable from a practical perspective with an average effect of 15.8 minutes when accounting for other variables and 30.1% difference in reported night wakings (though of note the range of averages in the SID group was approximately .65 to 2.2 with the peak of just over 2 happening at 5-6 months with the average onset of solids).

What Does This Mean?

I think the first issue we have to consider is whether we’re seeing much of anything happening here.  The differences are small enough to question whether we should consider these results “practically significant”.  With an absolute difference of 7 minutes based on average differences in the two groups (or even up to 16 minutes with other variables accounted for), it is difficult to see this as making that huge of a difference, especially not at the known costs of early introduction of solids.

The second issue is to do with what the ITT versus per-protocol analyses tell us.  Given that we see a stronger (though still questionably practical) effect with those who introduced allergens (and likely more protein given the sources), the question about what is driving the effect is up for debate.  Is it the protein content in the infant’s diet as protein will be harder to digest?  Is it that there is more solid food in general being given to those infants that were taking in the allergenic food?  Further exploration is needed.

The third issue is about how parents can interpret this.  Given the protocol adherence was so low for the EIG group, perhaps there is something about the actual children who were able to adhere that influences the results.  This raises the question of whether the effects can be replicated should more parents try to take the intervention to heart and introduce solids earlier.  This is one of the issues that arises when clinical trials protocols are used for interventions that are readily accessible – the ITT analysis only informs on the actual possible benefits of an intervention period whereas the per-protocol analysis doesn’t provide enough information on why less than half of the EIG adhered to the protocol given.  The why here is very important for the applicability of this to other families.

The final issue, and in my mind a very important one, is that the EIG parents reported far fewer sleep problems than the SIG parents and this was related to overall parental well-being.  Given that the reported sleep duration and wakings were actually quite similar, it lends support to the idea that – as mentioned above – the presence of an intervention that had parents doing something that has cultural support even if there’s little evidence for it (solids for sleep) influenced their perception of their child’s sleep.  Perhaps knowing they have actually “done something” allows them to better accept the wakings and the total sleep duration.  However, as the lower reporting of problems had a strong relationship to parental well-being, we should be looking to see if other interventions that don’t include feeding solids so early could similarly influence parental well-being.

Overall, it feels that this study does more harm than good.  The one positive is the one area completely ignored and that is the potential effect of an intervention on parental perceptions of problems and the subsequent possible link to overall well-being.  I would hope parents can take this research with a grain of salt and think about the larger picture and the minimal effect before trying a “cure” for a problem that doesn’t likely exist.


If you are in need of individualized parenting help, I offer services via email, Skype, and phone on a variety of parenting topics.  You can find out more here.