RE: "OBJECTIVELY MEASURED SLEEP CHARACTERISTICS AMONG EARLY-MIDDLE-AGED ADULTS: THE CARDIA STUDY"

Abstract

In the July 1, 2006, issue of the Journal, Lauderdale et al. (1) make an important contribution to a small, but growing literature on the social determinants of sleep (2–4). They analyze and present the objectively measured sleep characteristics of early-middle-aged adults in the CARDIA study in a clear and compelling manner. Specifically, they find that Black men and women spend less time in bed, get less overall sleep, sleep less efficiently, and have higher sleep latency than their White counterparts. Near the end of their article, they suggest that “sleep may prove to play a role in the associations between socioeconomic status, race, and health” (1, p. 15). This suggestion implies that short (and inefficient) sleep patterns by Blacks compared with Whites may explain part of the health inequalities widely observed between the two groups. At face value, this hypothesis is appealing because of the intuition that short sleep is bad for one's health. While we agree with the suggestion by Lauderdale et al. (1) that sleep may be related to racial inequalities in health, we do not agree with their methodology. In particular, the linear regression modeling they used implicitly assumes a consistent relation between length of sleep and health outcomes (i.e., longer sleep duration is associated with better health outcomes). This assumption overlooks widely observed findings that “overly” long sleeping is associated with increased morbidity and mortality (5–12). That is, although Lauderdale et al. find that Blacks may be sleeping less on average, it does not necessarily follow that they are worse off because of it. In fact, given that many studies find that long sleeping, compared with short sleeping, is associated with higher risks of adverse health outcomes (5, 8–12), another possible interpretation of Lauderdale et al.'s results might be that sleep duration is a mediating variable working to mitigate racial disparities in health. Rather than leaving the ambiguity in the interpretation, we argue that studies investigating the relation between sociodemographic characteristics and sleep duration should estimate a multinomial logistic regression in which short (<6.5 hours), midrange (6.5–8.5 hours), and long (>8.5 hours) sleep durations are the categorical dependent variables (2, 13). This simple change in methodology is particularly insightful in light of results presented by Hale and Do (13) at the annual meetings of the Population Association of America in April 2006. Using a population-based sample of US adults from the National Health Interview Survey, we showed that Blacks, compared with Whites, are significantly more likely to be both short and long sleepers, both of which are associated with higher health risks. That is, Blacks are more likely than Whites to sleep at both of the suboptimal ends of the sleep duration spectrum. This finding suggests that the linear regression model estimated by Lauderdale et al. (1) may be underestimating the amount by which Blacks are short sleepers because the long-sleeping Blacks may be pulling up the average for the short-sleeping Blacks. While Lauderdale et al. may be on to something good here, their methodology restricts them from identifying the full magnitude of the health risk related to the suboptimal sleep durations of Blacks. We strongly recommend that the authors reanalyze the data in a way that allows both of these high-risk sleep durations to be included. Conflict of interest: none declared.