An Unfounded Conclusion from a Confounded Study

To the Editor:

In the recent article by Edmonds and Vinson, “Three Measures of Sleep, Sleepiness, and Sleep Deprivation and the Risk of Injury: A Case-Control and Case-Crossover Study,”¹ the correlative relationships between self-reported sleepiness, sleep quality, and sleep amount with risk of accident-related injury are reported. The “major conclusion” is that “sleepiness and sleep deprivation, at least as they are commonly experienced, are weakly and inconsistently associated with injury risk.”

However, this conclusion is itself only weakly and inconsistently supported by the study findings: For example, the authors indicate that in their “multifaceted analyses of sleepiness and sleep deprivation, [there was] little evidence for association between [self-reported] sleep and injury,” despite also reporting that better recent sleep quality was associated with lower risk of injury. Similarly, based on a negative correlation between self-reported sleepiness and injury rate, the authors surmise that feeling sleepy might prompt behavioral changes that serve to reduce injury risk (ie, providing a “protective effect”). However, they also indicate that controlling for “location and activity” (the only 2 relevant aspects of behavior reported in the paper) had “little effect.”

Unfortunately, it would be difficult to draw conclusions from this study even if the findings had been consistent. This is because (as the authors themselves discuss) the results were not only subject to the considerable problems typically associated with retrospective self-report data, but the study was confounded: Injured subjects were interviewed in the hospital, under conditions that could reasonably be considered relatively stressful (eg, emotional turmoil associated with the recent accident), whereas control group subjects were interviewed by telephone (presumably at home, where they were less likely to be distressed). Compounding these problems, the authors failed to use any of the several available validated measures of subjective sleepiness (eg, Stanford Sleepiness Scale)² and sleep quality (eg, The Pittsburgh Sleep Quality Index),³ making it difficult to compare their results with those of prior field and laboratory studies.

Given the inconsistent findings, failure to use validated measures, and the confounding in this study, and especially given the fact that the authors were apparently aware of these problems (they acknowledge in the discussion section that the their results “might be due to confounding or reporting bias”), it was overstated to conclude that sleepiness may only weakly increase risk. Similarly, given the potential implications, the conjecture that sleepiness actually provides a “protective effect” against accidents by promoting more cautious behavior is not only unsupported by their data, it is contrary to the weight of scientific evidence.⁴

Even if their study had been well controlled, used validated measures of sleepiness, and produced consistent results, a more carefully considered conclusion would have been warranted. For example, the conclusion that “people involved in accidents resulting in injury are less likely to report having been sleepy at the time of the accident” would leave open the possibility that the injured subjects actually had been less sleepy, but also invites other, more plausible interpretations (eg, people involved in accidents resulting in injury are less willing to admit to sleepiness).