Post #4 Back-to-Sleep Series

Elizabeth Fama is back. This week, she explores the evidence linking sleep position and SIDS. It's an eye-opener.

In Post #1, Ms. Fama introduced and summarized her critical appraisal of the “Back to Sleep” campaign. In Post #2 she explored the possible downsides of supine sleep for infants. In Post #3 she gave us numbers on SIDS.

This week, Ms. Fama reviews the evidence linking sleep position and SIDS. It’s a tour de force. Readers of Sensible Medicine may notice many common lessons in this evidence review. JMM

How Strong is the Link Between Sleep Position and SIDS?

The number of recommendations for safe sleep quickly grew, based on findings at death scenes: in addition to sleeping the baby on her back, one should sleep the baby in her own crib or bassinette; with a firm mattress and tight sheet; without any blankets, plush toys, or bumpers (bumpers were federally outlawed by President Biden in 2022); near the parents but not co-sleeping; in a smoke-free home; in a cool room; and breastfeed her if possible. A relatively recent recommendation is to use a pacifier as the baby falls asleep, but not to force it on her and to leave it if it falls out, as well as new mentions of avoiding opioids and inclined sleep surfaces.

If the whole package has helped, how much is due to sleep position? Could we get much of the benefit of the program without the costs of back sleep?

Some studies try to tease out the separate causal impacts and mechanisms of these recommendations. For instance, smoking has multiple known biological effects on different systems in the infant’s body that have plausible connections with several proposed mechanisms of SIDS. A 2006 New Zealand article estimated that “assuming a causal association between smoking and SIDS,” 1/3 of SIDS deaths could have been prevented if their mothers hadn’t smoked. A 2010 American article estimates 23-34% of SIDS deaths are due to smoking.

In the popular imagination, SIDS strikes randomly, like an asteroid, and front-sleeping is the dominant risk factor.

As with many types of infant death, however, SIDS rates vary widely by intrinsic and extrinsic risk factors. The rate is higher for babies who are premature, of low birth weight, male, born into poverty, whose mothers received poor prenatal care, and whose mother or other household members smoked during pregnancy and/or after. Other risk factors include low education of the mother, young motherhood, single parenthood, unemployment, use of daycare (one study showed that 20.4% of SIDS deaths happen in daycare, rather than the 7% that would have been predicted by “extrapolating from census data”), older children at home, economically poorer geographical regions (true across countries, even in places like Japan with low SIDS rates), and disadvantaged racial groups. SIDS rates are higher with recent or current illness, and in winter months (December through February).

It’s thought that some percentage of SIDS deaths are due to infanticide. There are widely conflicting estimates of that number, ranging from 1.3% to 40% of SIDS cases. Suffocation without strangulation is impossible to distinguish from SIDS on autopsy—and obviously this is true for accidental suffocation as well.

SIDS is not the same asteroid for everyone. Whatever it is that we are calling SIDS is influenced by a range of intrinsic health factors of the infant, family traits, demographics, and baby-rearing practices. The conditional risk of SIDS varies a lot, and some of those conditions are controllable.

There Are No Randomized Controlled Trials

Early cross-country observational evidence implied that prone sleep was less safe: for example, it was well known in the 1980s that Sweden and Hong Kong had very low rates of SIDS compared with other countries, and that their cultural practice was to place babies on their backs. (Sweden has also always had a high co-sleeping rate, showing that co-sleeping can be done safely.)

Also, SIDS rates declined in country after country that initiated back-sleeping campaigns—drops that were apparently not due to reclassification of deaths.

However, the bulk of the evidence for the benefit of back-sleeping comes from hundreds of statistical investigations of the correlation between SIDS deaths and sleep position based on surveys of parents and caregivers of both SIDS and non-SIDS babies.

It’s an ocean of studies, each interpreting the data to conclude that prone sleep is less safe. There are only four exceptions, and they happen to be the earliest studies that included sleep position as a variable of interest: Carpenter and Shaddick (England, 1965), Froggatt in Bergman et al (Northern Ireland, 1970), Bergman et al (U.S., 1972) and Kahn et al (Belgium, 1984). Each of these says something on the order of “no conclusion can be drawn about the effect of sleeping position” (Kahn et al). In later papers, each of these primary authors comes around to supporting back sleep.

The apparent wealth of evidence and total consensus among researchers made me feel like a flat-earther for questioning the advice. Yet I found, when I examined paper after paper, that each is flawed, and the vast majority of them were researched and written after back-sleeping campaigns were in force (an important point I’ll return to). The primary problem is that all the studies are observational, ex post analyses, based on interviews of parents and caregivers. Even taken together, they do not represent high-level evidence—that which would be found in randomizing babies to sleeping prone or supine. Most researchers think, based on the consensus behind supine sleep, that a trial like this would be unethical.

(Editors’ note regarding this section: It was the multitude of observational studies that persuaded cardiologists to wrongly prescribe antiarrhythmic drugs for PVCs after MI, and hormone replacement therapy to post-menopausal women for cardiac protection.)

Two SIDS studies purported to be prospective, rather than retrospective, one done in Belgium in 1987-88 (published in 1990), and one done in Tasmania from 1988 to 1990. Both, understandably, suffer from low case numbers.

The Belgian study followed 4,064 infants from birth using monthly questionnaires and health visits. The 7 SIDS deaths that occurred during the first year were compared with 70 matched non-SIDS infants. The only significant differences were infant fatigue while feeding and a history of bedclothes soaked in sweat. Sleep position was certainly one of the 65 items they studied, but the authors never mention it. They conclude: “The limited number of SIDS cases collected….is probably the main reason why some of the epidemiological risk factors reported for SIDS….were not found.”

The Tasmanian study identified an at-risk cohort at birth, enrolling the parents in a program of one hospital interview and physical exam at 4 days, one home visit at 5 weeks (the only time “usual sleep position” was asked), and a phone interview at 10 weeks. They then analyzed the deaths that occurred over the first year. There are many problems with this study, beginning with the choice of enrolling only high-risk babies. There were 23 SIDS deaths out of 3,110 at-risk births between January 1988 and March of 1990, but only 15 had participated in interviews. The usual sleeping position (recorded at 5 weeks of age) was 6 on the side, and 9 prone. The authors seem not to have followed up on usual sleep position during the 10-week phone interview, which would have been helpful as 70% of the SIDS infants died after the 10-week mark. Finally, they had a survey strategy in which “if a control infant was unable to be interviewed, another appropriate infant was chosen,” which happened 17% of the time.

In an observational study, one takes a group of SIDS babies, and constructs a group of similar non-SIDS babies to serve as the control group. Early influential observational studies created their control groups by matching one or more characteristics: age at death date of the SIDS infants, sex, birthweight, community, hospital of birth, or in some cases just “50 most recent babies seen,” or “those responding first included,” or babies at their checkups, or in the UK by “same health-visitor list as SIDS victims.” Interviews regarding live infants were conducted variously by postal mail, via health visitors’ observations, at well-visits in clinic, and in home interviews. Many of the surveys and questionnaires were given to control parents at the end of a multi-year study—for example, surveys taken in 1984 for a study of SIDS by Dr. Susan Beal  covering 1970-1984—or sampled once in the middle of a years-long study period, or several weeks or months after the death of the SIDS case they were matched with. Parents of living children were asked to recall usual sleep position of their baby on the day of death of the matched SIDS victim, or during that month of life, etc. In one 1995 study in Norway, parents were asked about usual sleep position for infants up to 9 years prior.

Drawbacks of Surveys and the Challenge of Determining the Causal Contribution of Prone Sleeping

What we hope to know from all this research is the death rate for front sleepers vs. back sleepers, all else held constant.

But each part of this endeavor is difficult: death categorization is fallible, data on sleep position is surveyed ex-post, and “holding all else constant” is hard. Moreover, the “constants” are important to a family’s decisions.

Parents want to know, given our demographics, smoking and other history, given the health of our baby, given the other safe-sleep advice we do or do not follow, what is the risk of SIDS and how does it depend on sleep position?

Suppose we feel confident in our determination of which babies died of SIDS. Now we need to know not only how many of them were put to sleep on their fronts or backs, but also the sleep positions of all babies who did not die of SIDS. Research studies attempt to collect this last piece of data through population surveys described above. However, on the heels of widely disseminated sleep-position advice, when caregivers of both SIDS and non-SIDS babies are surveyed about their past practices, it’s possible that they report the “correct” answer, and the results will be mismeasured.

Of babies who die in their sleep and have a death scene investigation, parents and caregivers predominantly say that they placed the baby supine to sleep, yet the same data show that the unresponsive babies were more often found prone.

Hilina Kassa and colleagues studied 11,717 deaths between 2004 and 2014. They found that 42.4% of caregivers said they placed the child to sleep supine and 22.5% said prone, but 28.2% of the infants were found on their back and 39.7% were on their stomach. (The percentage of caregivers saying they placed the baby on their side or in an “unknown” position did not differ much from the percentages of infants found dead in those positions.) Kassa et al postulate that the infant may have rolled from the back to the “less safe” tummy. This is a problematic theory from a couple of angles: first, the medical intervention we’re testing is to put babies to sleep on their backs, not to turn them over again if they roll onto their stomachs (even the AAP advises allowing tummy-rollers to sleep as they prefer); second, I’ve shown that back-slept babies are less able to roll over before six months, the age at which SIDS risk quickly declines (72% die between 1 and 4 months, and 90% die before 6 months, per NICHD). Moreover, American babies tend to learn tummy-to-back rolls first. It seems more likely that the caregivers who report the baby’s sleep position are misremembering in the fog of grief or are reluctant to admit to the “mistake” of placing the baby prone. Another possibility not considered by Kassa et al, but mentioned in older articles like Carpenter and Shaddick 1965, is that “agonal movement may, of course, affect the position in which the child is found.”

On its own, the Kassa et al data above might suggest to you that prone sleep is even more dangerous than expected: babies reported as dying on their backs were really on their stomachs. But SIDS is a rare event, and these data only tell us, of babies who died, what fraction were placed or found prone vs. supine? The question we want to know is, of babies who are placed prone or supine, what fraction die?

As an extreme example: if 3 out of 4 babies who died were found on their stomachs, it seems to imply that tummy sleeping is dangerous. Yet suppose that among all babies, 9 out of 10 were sleeping on their stomachs. Now we would conclude that prone sleep is in fact safer than supine. More of the babies who were sleeping on their backs died. If parents skew their answers toward supine sleep, we get a biased measure of the dangers of prone sleep. The paper by Kassa et al seems to show that this bias exists.

In looking at the time trends of Figure 1 (in Post #1), then, to some extent we could simply be measuring whether parents know the right answer. Similarly, when we look across states or countries, we may not be seeing solely the variation in tummy or back position by locale, we could also be seeing a differing awareness of public health messages.

Yet another potential problem with observational case-control SIDS research that I have rarely seen discussed (with one exception here) is whether “sleep-positioning action” (my term) is reported for SIDS and controls. “Usual position,” “position placed in,” or “position found in” are all different things. Some studies use the ambiguous phrase “sleeping position,” and there’s no way of knowing which of the three (usual, placed, found) they mean. In a randomized experiment, the intervention would be to put a child to sleep on her stomach, side, or back, as that is the action we should care about. In some influential studies, for instance Susan Beal 1986, deceased babies “found” prone in Australia are compared with “usual position” of infants in other countries, which could be apples and oranges. A SIDS baby who had an agonal event could have a different “found in” than “placed in” or “usual position.” A baby capable of rolling may be placed to sleep in one position but routinely choose another every night.

Similarly, it’s rare for studies to report the position of the baby’s face when found prone: with face to the side or face-down in the mattress. Hunt and Shannon noted this in a comment in Pediatrics in July of 1992.

American researchers were late to the game, conducting observational studies after scientists in Australia, U.K., New Zealand, Netherlands, Hong Kong, and France had already homed in on prone sleep as a culprit in cot death.

Gilbert et al (2005) ran a large systematic review of SIDS studies related to sleep position, and the first U.S. study on their list occurred in 1992 by Hoffman and Hillman. Recall that 1992 is the year AAP stressed back and side sleeping only, based on the precedent of overseas research, so the first American scientists may have already been influenced by those guidelines when designing their studies and analyzing the data.

Surprisingly, Hoffman and Hillman seem mildly equivocal about the sleep-position results they get using a very large dataset gathered by the NICHD Cooperative Epidemiological Study. They report that 81% of SIDS babies “usually” slept prone, compared with 77% of the matched control cohort, a relative risk to prone sleep of 1.3, with a p value less than 0.05.

"Of course, many other risk factors were found to be more strongly related to SIDS in the NICHD study,"

they say, but they also expect that:

“Within the next 2 to 3 years, much stronger evidence should be available to determine whether prone sleeping position among infants in Europe and the United States is an important risk factor for SIDS.”

Professional recommendations were far ahead of them: by 1988 in the U.K., Gilbert et al could not find a single infant-care book, manual, or pamphlet that still recommended prone sleep. Even Doctor Spock, published in the U.S., had changed his advice by 1988.

Side-sleeping, initially found to have odds ratios hovering around 1 (meaning, no more likely to cause SIDS)—which I assume is what caused the AAP to continue to allow the position until 2005—began in the late 1990s to have odds ratios as high as 5 (see this 1998 Netherland’s study).

The timeline of this massive quantity of research and the persistent campaigning of a few early researchers suggest either a brilliant discovery (supine sleep) that was subsequently and resoundingly corroborated by hundreds of studies, or a groundswell of professional opinion that nudged all subsequent research to agree with growing public policy.

Or perhaps more likely, something in the middle: supine sleep has a small effect on SIDS for some, or a large effect for a few, but it may be unnecessary for the vast majority of healthy full-term infants (holding all other risk factors equal), while research has tended to exaggerate the benefit.

---

In the thumbnail image there are at least four violations of Safe-to-Sleep advice, but luckily this baby is now a thriving 33-year-old teacher, comic artist, and swing dancer.

Next week's post will conclude this series. Ms Fama argues that we still have more questions than answers, and asks whether the downsides of the Back to Sleep campaign disproportionately affect some groups.