SLEEP, Vol. 28, No. 9, 2005 INSOMNIA IS ONE OF THE MOST PREVALENT SLEEP DISORDERS WITH NEARLY ONE THIRD OF THE POPULA- TION HAVING AT LEAST MILD FORMS OF INSOMNIA. 1 Even when using the most restrictive diagnostic definition, there is still about 6% of the population who has insomnia. Asking if insomnia kills is certainly a legitimate question. The paper by Philips and Mannino is an elegant attempt to answer that question in a large community sample. Using a 6.3-year follow-up, the au- thors demonstrated that insomnia complaints and use of hypnotic medication were not associated with an increased mortality after controlling for other key variables. This is an important finding. However, the answer to that question is complex. The publica- tion by Kripke and his colleagues in 1979 2 showing that short (less than 4 hours) and long sleep (10 hours or more) were associ- ated with increased mortality, was a wake up call to the scientific community regarding to mortality risks associated with insuffi- cient or excessive sleep. Unfortunately, research in this area is still insufficient. About a dozen of scientific articles have addressed this issue with inconsistent results: increased risk ratios for mor- tality were found in subjects who reported insomnia or abnormal sleep patterns 2-5 or used sleeping pills 2,3,6,7 while other studies did not find insomnia to be related to increased mortality. 7-10 Several aspects need to be clarified before any definite conclusion can be reached. The first difficulty to overcome is how to define insomnia in the general population. As previously underlined, 1,11 a large defi- nition will include many individuals with dubious insomnia and will prevent to reach reliable conclusion. Insomnia encompasses more than the mere presence of sleep onset or sleep maintenance difficulties: it has a value only when it has a certain frequency, duration or recurrence and when it somehow impairs the func- tioning of an individual. Philips and Mannino made a commend- able effort to narrow the definition of insomnia in their study by combining the presence of difficulty initiating or maintaining sleep with non-restorative sleep. It is a clear strength compared to other mortality studies. Regrettably, frequency and duration of insomnia were not assessed. Therefore, their group of subjects with insomnia is heterogeneous including mild to severe insom- nia and transient to chronic insomnia posing the problem of large within-group variability. This limitation was observed in nearly all mortality studies. The problem is that insomnia mortality risk studies often derived from surveys that were designed for other investigational purposes. Consequently, the number of questions addressing sleep issues was most of the time limited to 1 to 4 questions on insomnia symptoms and, sometimes, sleep duration was also asked. The time frame (past month, past year, lifetime) and frequency were rarely delineated and the duration of insom- nia absent. The second difficulty is to establish causality between insom- nia and mortality causes. Several risk factors frequently associ- ated with insomnia are in some instances a cause for insomnia and in other cases they are a consequence. One example is poor health status. Degradation in health may lead to insomnia as the appearance of insomnia may contribute to deteriorate health. 12 Therefore, the question is how often insomnia has triggered or has preceded the condition that led to death. To date, this question remains unanswered. Furthermore, we know little about the role of insomnia in the development of several diseases. For example, cardiac disease, a leading cause for death, has been several times associated with insomnia. However, studies assessing the influ- ence of insomnia on the appearance of cardiac disease had the same limitations than mortality studies (inadequate definition of insomnia and lack of control for confounding factors). 13 The third difficulty is in selecting the confounding factors to introduce in the predictive model. The selection should be based on several considerations: 1) their significant association with the studied outcome or their biological relevance (for example, age and gender); 2) their frequency in the sample and; 3) their minimal collinearity with the other confounding variables (ie, information redundancy). A consequence in introducing collinear variables in a model is that the variables can nullify each other, mask signifi- cant associations and allow the emergence of spurious risks. The Philips and Mannino study is notable for the carefulness the au- thors took in defining other risk factors like hypnotic use, alcohol intake, smoking status and lung function. These variables have a greater precision compared to other sleep and mortality stud- ies. This is important since it reduces within-group variability for these variables and strengthens the results. To date, the evidence linking (or not) insomnia with increased mortality risks remains inconclusive. It is imperative that future studies improve the assessment and definition of insomnia in or- der to identify different groups of subjects with insomnia such as transient and chronic insomnia and the severity of insomnia. The evaluation of severity should include frequency of symptoms and impact on functioning. Longitudinal studies that have assessed insomnia on different occasions have the possibility to evaluate if subjects with persistent insomnia (ie, present at least at 2 different assessments) have a higher mortality risk. In the eyes of the population, the current scientific knowledge about insomnia and mortality risk may appear somehow baffling. On one hand, studies show that you may die younger if you have Insomnia: A Dangerous Condition but Not a Killer? Maurice M. Ohayon, MD, DSc, PhD Stanford University School of Medicine, Stanford Sleep Epidemiology Research Center, Palo Alto, CA Editorial—Ohayon Disclosure Statement Dr. Ohayon has received research support from Lilly; and has received hon- oraria from Lilly and Excerpta Medica. Address correspondence to: Maurice Ohayon, MD, Stanford Sleep Epidemi- ology Research Center 3430 W. Bayshore Road, Suite 102, Palo Alto, CA 94303; Tel: (650) 947-9812; Fax: (650)947-9813; E-mail: mohayon@stanford. edu 1043 EDITORIAL Comment on Phillips B; Mannino DM. Does Insomnia Kill? SLEEP 2005;28(8):965-971