Sunday, December 14, 2014
SIDS, Outcome Research, and Proxy Measures
I mentioned a few days ago that Ed Clint of Skeptic Ink had objected to my comments on the research evidence behind the Back to Sleep/Safe to Sleep campaign to have parents put young babies in the supine position for sleep. Without concluding that a particular position was safest for infant sleep, I had queried whether the epidemiological research available actually provided evidence that supine sleeping reduces the rate of Sudden Infant Death Syndrome, as has been claimed. Now Ed has posted his own take on the situation, at http://www.skepticink.com/health/2014/12/11/back-sleepsafe-sleep-saveschildrens-lives/.
In his post, Ed included a familiar graph that shows 1.4 SIDS deaths for every 1000 live births in 1988, a gradual decline to 1.2 SIDS deaths per 1000 in 1992, the first year of the American Academy of Pediatrics recommendation for supine sleeping position, a reduction to 1.03 per 1000 in 1994, the first year of the Back to Sleep campaign, and then a continuing gradual reduction to a rate of 0.55 per 1000 in 2006. The graph also indicates changes in proportions of parents using prone or supine sleeping positions for their infants. From 13 % sleeping supine in 1992, this figure gradually rises to 75% in 2006.
This graph looks pretty significant, doesn’t it? Ignoring the fact that the rate of SIDS deaths was already decreasing before any recommendations for supine sleeping occurred, we see that the increase in the proportion of babies sleeping supine is accompanied by a decrease in SIDS deaths. It’s easy to assume that one caused the other… but of course the graph doesn’t show everything that could be affecting the SIDS rate. What other factors should we be looking at?
There are quite a few factors other than sleeping position that are associated with a lower or higher SIDS rate. Let’s just look at one of them—a variable about which much research has been done : tobacco smoking. Maternal smoking during and after pregnancy is known to be associated with a higher risk of SIDS. So, what has been happening about smoking during this period when the SIDS rate has been dropping? You can see this on slide 7 at www.cdc.gov/winnablebattles/tobacco/ppt/tobacco_wb_presentation_-oct2014_rev_508compliant.pptx. For those readers who hate trying to read graphs, I’ll describe what it shows: Between 1990 and 2010, a period roughly equivalent to the time shown on the SIDS and sleeping position graph, there was a reduction in cigarette smoking in the United States, from 3000 cigarettes per capita per year in 1990, to about 1000 in 2010. If we were to draw a new graph showing the relationship between SIDS and smoking, it would look very much like the association between supine sleeping and SIDS (only reversed, so that less smoking goes with less SIDS, whereas less SIDS goes with more supine sleeping, less prone sleeping).
Assuming for the sake of argument that SIDS is caused primarily by one of these two factors (which might not be the case), we apparently need to ask which one it is. The graphic representation of the relationships between SIDS and each of the factors look just alike. If we are going to have a campaign to reduce the SIDS rate, and we are going to tell people to do just one thing, which will it be-- don’t smoke, or don’t have the baby sleep prone? The figures don’t, and indeed can’t, answer this question, and can’t even tell us whether our two candidate factors are the only important ones. The choice is made for other reasons, and this is why I said that this kind of policy is in fact not strictly speaking science-based.
But let’s look a bit at the issue of whether either smoking or sleeping position is actually at work in affecting the SIDS rate. The problem with epidemiological research is that the factors looked at may be only proxy measures that are markers for the existence of some other factor that is more difficult to measure. Take, for example, a study that looked at (among other things) the effect of socioeconomic status on SIDS rates (Gelb, Aerts, & Nunes (2006). Sleep practices and sudden infant death syndrome: A new proposal for scoring risk factors. Sleep, 29, 1288-1294). Rather than do an elaborate and intrusive investigation into a family’s income and property to decide SES, these authors used maternal education as a proxy measure for SES, assuming that high maternal education is more likely among affluent families. Low maternal education was associated with the use of a set of high-risk sleeping practices, including not only prone position but soft bedding and allowing the child’s head to be covered during sleep. Young mothers and smoking mothers were more likely to use high-risk sleep practices, as were mothers who had fewer prenatal doctor visits. The mothers’ practices and their other characteristics were confounded in such a way that it was difficult to know what factor caused any outcome.
Proxy measures can be useful shortcuts to measurements that are hard to do. However, it’s important to remember that any factor chosen may in fact be a proxy measure that is determined by some background factor that is unnamed. When parents choose to comply with advice about supine sleeping, is the factor that influences SIDS sleeping position itself, or is this just a proxy marker for parental characteristics such as carefulness, compliance with rules, middle-class status, education, and affluence? As developmental psychologists have proposed with respect to other topics, it may be that “good things go together” and that we don’t know which “good thing" is at work.
Choosing one “good thing” to work on may allow a simple campaign to be developed, but, once again, that choice is not based on clear scientific evidence about SIDS.
N.B. The point here is not to advise about infant sleeping positions, but to discuss the complexities of extracting helpful advice out of the available information.