Copyright: Kirill Federspiel

Copyright: Kirill Federspiel

Bedsharing remains one of the more controversial decisions a parent can make these days, with articles touting that doing so will result in you killing your child, sharing horror stories of families who woke up to a dead baby, all while leaving out any information about risk factors or other environmental considerations.  The American Academy of Pediatrics[1] and their doctors continue to speak out against the practice, leaving no room for families to speak to doctors about safety concerns unless the doctor is willing to go against the AAP.  Despite this, some researchers and medical professionals have spoken out against this perceived “no bedsharing” message in favour of a more nuanced message about safe bedsharing and the individual circumstances that families find themselves in.  For example, in the UK, the “no bedsharing” message is being revised by some health agencies in favour of focusing on education about safe bedsharing and nuanced discussions about each individual family’s needs and risks[2].

The question that many people ask is: How can two different sets of professionals read the research and end up with totally opposite conclusions?  Why does the governing body in the USA feel so differently from that in the UK (or elsewhere, like Japan, where bedsharing is incredibly common)?

For years, everyone thought bedsharing was dangerous based on a slew of epidemiological studies looking at risk factors for infant deaths – specifically SIDS or suffocation.  In those studies, bedsharing was one of the main risk factors of death and so it seemed prudent for most governing agencies to take an official stance against it.  However, things got muddier as time went on and research progressed.   Newer studies weren’t so clear-cut, with other factors – like smoking or drinking – interacting with bedsharing to increase the risk of harm to an infant.  And like a pendulum that swings back and forth, when one study emerges that seems to clear bedsharing, another hits the press that seems to vilify it.

In recent years, two large studies seem to be oft-quoted in support of this anti-bedsharing stance: The first is a meta-analysis by Carpenter and colleagues[3] and the second is a study that found a higher risk of SIDS/suffocation associated with bedsharing in younger infants (0-3 months) than older infants[4].  Other studies that have come out at similar times (e.g., [5]) but come to opposite conclusions, do not receive nearly the same press.  It’s as if our society has made up their mind and the media is playing part to a confirmation bias about the dangers of bedsharing.

Does the research really support an anti-bedsharing message?  What should parents really be taking home?  My hope here is that by looking at the research that is being used to form opinions on bedsharing, I can show you what is involved in the research and allow you to come to your own conclusion about the real risks inherent in bedsharing versus the risks of other factors interacting with bedsharing.

Given the focus on the larger meta-analysis conducted by Carpenter and colleagues recently, I have decided that the main studies that were included in this analysis should be the primary focus of this discussion as they are what seems to be informing people’s opinions.  I also include some newer studies that also seem to suggest bedsharing is problematic as well as those that seem to not find the same level of risk demonstrated in the aforementioned studies.

The Risk of Bedsharing Per Se vs. Other Factors

Source: ISIS Co-Sleeping Image Archive

Source: ISIS Co-Sleeping Image Archive

When the first bits of epidemiological research emerged, bedsharing was found to be a large factor in contributing to infant death, something that was mirrored by Carpenter and colleagues in 2013 (as these were the studies that were included in this meta-analysis).  These studies seemed to find a 2-3 times increased risk of death associated with bedsharing[6][7][8][9].  Although the recent meta-analysis boasts that it is a meta-analysis, giving us the best possible answer to the question of risk in bedsharing, the data we obtain from any meta-analysis is only as good as the data that goes into it.  In this case, many variables were omitted or missing data from so many cases that the meta-analysis should be interpreted with extreme caution.

Looking at the individual five, epidemiological studies included, we see the following missing factors:

  • The New Zealand study [6] failed to account for alcohol use, type of bedding and tog value, and sleep surface (e.g., bed versus sofa). However, as this was one of the first studies done on the topic, it is difficult to fault the researchers for failing to assess these variables.  The later studies should fall under greater scrutiny given that they had prior knowledge of factors that interacted with bedsharing to increase the risk.
  • The European case control study [7] did include many of the confounding factors, including alcohol use and duvet use, but not all centres reported on these variables, thus the data is incomplete. Notably, only 78% of the places reported on alcohol use and 56% reported on duvet use; the rest was predicted using regression, a method that has since been found to be flawed as it tends to overestimate significance.  In this case, this would mean that alcohol use and duvet use were overpredicted and depending on whether more missing data was present in the control versus risk groups, this could sway the results either way.
  • The Scottish study [8] omitted so many variables it’s difficult to know how to interpret the findings. The study failed to account for alcohol use, duvet use, breastfeeding (though they had data on breastfeeding), smoking during pregnancy, and illegal drug use.
  • The Irish study [9] actually collected the most data and included a lot of important variables such as tog value of clothing plus bedding, baby prone to sweating, smoking, drinking, and more. They also assessed whether bedsharing was the usual routine or not, though did not include this in the adjusted odds ratio for bedsharing despite finding that for 50% of the SIDS cases that were found bedsharing, this was not the usual sleep pattern.  Three large interactions with bedsharing were found in this study: Maternal smoking, birth weight adjusted for gestation, and tog value.  Indeed, in this study, when looking only at non-smokers, ignoring the other two interactions with birth weight and tog value, bedsharing was not statistically significant as a risk factor.
  • The German study[10] poses a slight problem for someone like me because the key variable missing in this report is bedsharing. Yes, according to the citation Carpenter and colleagues provided, there is no information on bedsharing at all.  This does not mean that this data was not collected and provided to Carpenter, but that we have no way of knowing what the risk was or how it interacted with other variables.  As such, I can make no comment about the risk of bedsharing in this particular study or what variables were included or omitted.

Thus, by the time the meta-analysis was performed, over 60% of cases were missing information on alcohol use and illegal drug use, whereas no data on type of bedding or tog value (e.g., the use of a duvet) and prenatal smoking were included in the analyses, despite these being large risk factors identified in the original research.  For example, smoking prenatally was associated with a 4.09x increased risk of death in [6], duvet use was found to be associated with a 2.95x increased risk of death in [7], and tog value > 9 in [9] was found to increase the risk of death by 9.68x, all larger than the risk of bedsharing.

Defining Variables, or ‘The Variable Problem’

The variable problem is one that is rampant in a lot of research, especially parenting research.  I have discussed it with respect to various parenting methods researched today, such as breastfeeding and sleep training, and it sadly also plays a role in the bedsharing research.  The variable problem refers to the fact that many variables, as they are used in research, are poorly defined and this influences the types of conclusions we can make from any research.  In the bedsharing debate, there is little concern over the definition of bedsharing (though there is concern about reporting which I will touch on below), but how some of the confounding factors are defined and measured can greatly influence how much of a role they are deemed to play.

On top of the issue of missing variables, here are some of the ways in which the major case control studies may be subject to the variable problem:

  • One of the glaring examples from the New Zealand study[6] was the way in which breastfeeding was coded. Namely, instead of looking at a dose response or even current breastfeeding, the researchers included breastfeed as positive if the mother engaged in “any” breastfeeding.  This fails to capture the information needed to properly assess the real risk, especially as we know that not breastfeeding does increase the risk of SIDS significantly.
  • The European case control study[7] also suffered from the variable problem. In this case, alcohol was measured as the number of drinks the mother had consumed in the previous 24 hours using groupings of none, 1-2, and 3 or more, ignoring the possible real effect of timing on alcohol use.  To clarify, a mother who has a glass of wine at lunch is now lumped in with the mother who has two shots before bed.  This is likely why the authors found that 1-2 drinks was not associated with any increased risk compared to no drinking at all.  Timing is, after all, everything.
  • In the Scottish study[8], the key smoking variable was defined as “either parent smoked” meaning that the lower risk for a partner smoking was now confounded with that of the mother smoking. This means that women who were not smokers were now included as having this risk when they didn’t.  This has the potential for lowering the odds ratio for the smoking by bedsharing interaction as a risk factor.  In addition, breastfeeding (though not included statistically in the model and only examined on its own) was recorded as “any breastfeeding”, again missing out on the dose response and the interaction with bedsharing.
  • The Irish study[9] continued with the problems in defining breastfeeding, which was simply defined as “breastfeeding initiated at birth” or not. Again, as with European case control study, parental alcohol consumption was measured over a 24-hour period which may explain why the variable was non-significant.
  • The German study[10], as reported, did not include bedsharing and it also did not provide information on other variables of interest for this examination. Instead the focus was on medical conditions that may have explained the infant’s death and although this is highly relevant for the discussion, it does not speak to the control of other risk factors that interact with bedsharing, such as alcohol consumption by parent or breastfeeding.

In addition to the problems with the variables themselves, there is the issue of reporting.  The concern here is how accurate the reporting of bedsharing is in the control group.  Why would we think there is a problem?  Well, anecdotally many parents admit to underreporting bedsharing when they are faced with questions by health professionals, especially when the presiding view is that it is “wrong” or “bad”.  The rates of bedsharing in the control cases or when asked by a doctor do not mirror studies which ask in a more benign or anonymous manner.  To highlight the discrepancies, research presented at the Pediatric Academy Societies meeting in 2014 reported 20% of parents “regularly” bedshare, using a large sample of women recruited from hospitals.  However, another study found that 72% of parents of 3-month-olds reported “regularly or occasionally” bedsharing with their infant[11].

Now, a couple differences in these studies: 1) The first was conducted in the USA whereas the second was in Canada; and 2) the first looked only at “regular” bedsharing whereas the second included “occasional” bedsharing as well.  This second point brings up another issue: The frequency of bedsharing.  For the sake of this discussion, frequency is an important factor as many of the dangers may come from “occasional” bedsharing, especially if it occurs unplanned and due to immediate circumstances.  Returning to the Irish study that was included in the Carpenter analysis[9], the degree of regularity of bedsharing was found to be irregular in the majority of deaths where bedsharing was a factor.

Additionally, in most studies the control group reports only on the previous night.  However, if a vast majority of families only “occasionally” bedshare, the previous night may be missing important information on risk.  Another factor is that many families only bedshare part of the night and yet report that they do not bedshare, often because they didn’t start the night there or end the night together.  This hides the fact that bedsharing is taking place and thus artificially raises the risk in the research.  Additionally, if people are not regularly bedsharing in the case group, then the chances that they are doing so safely and without other risk factors is dramatically decreased.

Is Age Everything?

Source: ISIS Co-Sleeping Image Archive

Source: ISIS Co-Sleeping Image Archive

One of the key findings that is consistent across any study that finds a risk of bedsharing is that it is age-dependent.  It is pretty conclusive now that bedsharing after 3-4 months of age is not associated with any increased risk, no matter how the research is done, but the debate around younger children remains.  In line with this, the other recent study that made headlines was an analysis of risk factors for infant death by age[4].  This study was interesting in that it found more younger infants died in bed with a parent than older infants did, but omitted a key variable: the base rate of bedsharing.

Why is the base rate important?  Imagine if you had two groups of students and you wanted to know how they differed on how many students didn’t do their homework, measuring it as a yes/no of whether or not they turned in homework each day.  Would this be a valid measure?  Hopefully you can see it’s not because what we’re missing is the base rate of who has homework each day.  That is, if one group is more likely to have homework, they are also more likely to do it and turn it in, resulting in an erroneous comparison between the two groups.

A similar thing occurred in this study in which the actual base rates for bedsharing between the two groups were not examined.  That is, one needs to look at the proportion of infants who bedshare at the various ages and make comparisons to those rates, not comparisons across ages.  Thus, although younger infants were more likely to be found bedsharing, it is entirely plausible that parents are more likely to bedshare with younger infants more generally.  If this is the case, then bedsharing is not a risk, but the sleeping location simply reflects where infants normally sleep.  In fact, another, more recent study examined this very issue in younger infants while controlling for the known risk factors and found no significant increased risk even in younger infants[5].

Is One Death Too Many?

The final argument that is often made is that “one death is too many” in terms of bedsharing.  The assumption here is that all bedsharing deaths would be prevented if babies slept in cribs and no crib deaths would be prevented by bedsharing.  When bedsharing is done unsafely, then yes, it would likely (though not always) be prevented by using a cot safely; however, this speaks to education on how to safely bedshare instead of condemning the entire practice.  It is unlikely that, absent known risk factors, an infant bedsharing would be safer in a crib, and this is exactly what more comprehensive research has found: No increased risk, but this does not mean no risk.

This issue also speaks to the etiology of SIDS which is looking more to be biological in nature (though it may interact with environmental factors).  If the etiology of SIDS is linked to neurological disfunction, then sleeping location may not matter much at all.  In fact, we may have reverse causality at play.  That is, if a child has a predisposition to SIDS, it may manifest in ways that affect sleep (e.g., sleep apnea, fussiness/crying) and this may lead parents of these children to bring their babies to bed with them in order to either watch over them or try to get some sleep themselves.  In this case, the bedsharing could be the result of the problem that leads to SIDS.

As we don’t know the actual etiology of SIDS, but we know that suffocation can be a risk factor and that bedsharing rates are high and rising, this begs the question: Why are we not teaching parents safe bedsharing?  Would these deaths be avoided if we were able to help families prepare their environments in a safe manner and recognize their own risk factors when it comes to bedsharing?

Conclusions

I used these recent studies to highlight the problems with bedsharing in the literature and to explain why some groups are moving away from a strict interpretation of the research as suggesting “no bedsharing”.  There are other studies out there that suggest bedsharing is dangerous, but I have yet to come across one that did not suffer from the main problems highlighted herein, namely either omitting key variables or having ill-defined variables that limit interpretability.

For example, recent research in the American Journal of Public Health[12] which condemned bedsharing due to the finding that a majority of SUIDs were found in an adult bed, yet the authors did not include any assessment of breastfeeding, tog value, alcohol consumption, smoking, and so on.  As the focus of this examination was strictly the sleep environment, this made sense for their purposes, but fails to consider the facts we know interact with bedsharing, making the recommendations against bedsharing outrageously premature.  [Of note, a higher proportion of SIDS cases (SUID encompassed SIDS, suffocation, and undetermined cases) were still found in a crib (37.7%), not an adult bed (33.5%), but again we have no baseline values to compare to making any conclusions impossible.]

You may read this and feel that bedsharing is still something that is inherently dangerous, though I hope that after reading this you are able to see how the limitations in our research really do influence how we’ve taken data that is problematic and used it to make recommendations that fail to capture biological needs, human nature, and human history.  The question has to become, how will we best benefit families?  Is a blanket ban the way to go or would more education and information (while opening up lines of communication with health care providers) be more effective?

If you are considering bedsharing, please check out this amazing book by Dr. James McKenna which reviews the safety issues surrounding bedsharing and gives you tips on how to do it safely (image links to Amazon as an affiliate link):

Sleeping with Your Baby: A Parent’s Guide to Cosleeping

You can also check out this abridged version:

___________________________

[1] Task Force on Sudden Infant Death Syndrome.  SIDS and other sleep-related infant deaths: expansion of recommendations for a safe sleep environment.  Pediatrics 2011; 128: e1341-67.

[2] http://www.nice.org.uk/guidance/CG37

[3] Carpenter R, McGarvey C, Mitchell EA, Tappin DM, Vennemann MM, et al. Bed sharing when parents do not smoke: is there a risk of SIDS? An individual level analysis of five major case-control studies.  BMJ Open 2013; 3: e002299; doi: 10.1136/bmjopen-2012-002299.

[4] Colvin JD, Collie-Ackers V, Schunn C, Moon RY.  Sleep environment risks for younger and older infants. Pediatrics 2014; doi:10.1542/peds.2014-0401.

[5] Blair PS, Sidebotham P, Pease A, Fleming PJ.  Bed-sharing in the absence of hazardous circumstances: is there a risk of Sudden Infant Death Syndrome? An analysis from two case-control studies conducted in the UK.  PLoS One 2014; doi: 10.1371/journal.pone.0107799

[6] Mitchell EA, Taylor BJ, Ford RPK, Stewart AW, Becroft DMO, et al. Four modifiable and other major risk factors for cot death: The New Zealand study.  J Paediatr Child Health 1992; 28: S3-8.

[7] Carpenter RG, Irgens LM, Blair PS, England PD, Fleming P, et al. Sudden unexplained infant death in 20 regions in Europe: case control study.  The Lancet 2004; 363: 185-91.

[8] Tappin D, Ecob R, Stat S, Brooke H.  Bedsharing, roomsharing, and sudden infant death syndrome in Scotland: a case-control study.  J Pediatr 2005; 147: 32-7.

[9] McGarvey C, McDonnell M, Hamilton K, O’Regan M, Matthews T.  An 8 year study of risk factors for SIDS: bed-sharing versus non-bed-sharing.  Arch Dis Child 2006; 91: 318-323.

[10] Findeisen M, Vennemann M, Brinkmann B, Ortmann C, Röse I, et al. German study on sudden infant death (GeSID): design, epidemiological and pathological profile.  Int J Legal Med 2004; 118: 163-9.

[11] Ateah CA, Hamelin KJ.  Maternal bedsharing practices, experiences, and awareness of risks.  Journal of Obstetric, Gynocologic, & Neonatal Nursing 2008; 37: 274-81.

[12] Schnitzer PG, Covington TM, Dykstra HK.  Sudden unexplained infant death: sleep environment and circumstances.  Am J Public Health 2012; 102: 1204-12.