Longitudinal Data Are Needed

Studies on the possible influence of sleep facets on health and disease promise important information and causal insights (1, 2). Generally, this is also true for the cross-sectional study by Buchmann et al. regarding “Sleep, Muscle Mass and Muscle Function in Older People“ (3). Specifically, though, if we want to reliably judge whether observed associations are causal, longitudinal information should be considered. Initially, the authors avoid a reductionist approach when analyzing the role of “sleep” by considering several facets of sleep [including “sleep duration”, “sleep quality” and “sleep efficiency”]. However, that such study data are confined to cross-sectional information can lead to considerable problems when interpreting observations. As an illustration of this, recently, extensive meta-analyses of possible associations between facets of sleep such as “sleep duration, “napping” and a category of “poor sleep” and incident cancers were published (4). One key result was that observational studies must consider longitudinal information on sleep facets to take note of the complexity of multidirectional relationships between “sleep and the development of cancer”. Epistemologically, the insight that sleep disorders can be cause and/or consequence of disease over time should not be confined to the endpoint “cancer”. Indeed, this rationale could be relevant for possible causal relationships between “Sleep, Muscle Mass and Muscle Function in Older People”—as mentioned by the authors—as well. Buchmann et al. discuss that the different results in men and women can not be unambiguously clarified because auf the cross-sectional study design. Moreover, the authors conclude that longitudinal studies are necessary to deduce in what way muscle mass depends on sleep behaviour.

Overall, it cannot be emphasized too strongly that risks of diverse endpoints may be masked, but can also be spuriously elevated, in cross-sectional studies: Misclassifying variant sleep facets over years and decades as constant is a study limitation—not specified as such by the authors – which can disallow causal interpretations.