Additional related information may be found at:
Neuropsychopharmacology: The Fifth Generation of Progress

Back to Psychopharmacology - The Fourth Generation of Progress

Disordered Sleep :

Developmental and Biopsychosocial Perspectives on the Diagnosis and Treatment of Persistent Insomnia

Charles F. Reynolds III, Daniel J. Buysse, and David J. Kupfer


The maintenance of a robust circadian rhythm of sleep and waking is a vital component of successful adaptation across the life cycle, in that sleep and circadian rhythms help to regulate mood and govern the ability to perform cognitively. Changes in sleep, sleep quality, and daytime alertness have enormous impact on quality of life, level of functioning, and (particularly in late life) ability to remain independent (43, 45, 47, 70). Hence, preserving the integrity of sleep across the life cycle is a major public health priority and is, we suggest, a critical correlate of "successful aging" (59).

In this chapter we will focus primarily on persistent insomnia as a prototypical sleep disorder because of its importance to psychiatry and neuropsychopharmacology. The symptom of persistent insomnia can result from numerous factors operating singly or in combination. The most important factors contributing to insomnia include psychiatric disorders, behavioral patterns destructive to sleep (e.g., worrying excessively about sleep, irregular sleep–wake schedule, spending excessive time in bed), medication and substance use, medical/neurological illness, circadian dyssynchrony, and intrinsic sleep pathologies (25) (see Neuropeptide Alterations in Mood Disorders). However, considerable debate continues regarding the relative importance of these factors in the conceptualization and presumed etiology of persistent insomnia. Some investigators have argued that most insomnia complaints reflect psychiatric and psychological disturbances (32, 64). Other investigators have emphasized the importance of other causes, often relying on polysomnographic information to make causal inferences (11, 15, 30).

Within this context, the goals of this chapter are as follows: (a) to review epidemiological data on the prevalence of sleep disturbance, particularly of persistent insomnia, in relation to psychiatric disorders; (b) to review medical correlates of persistent insomnia in clinical samples; (c) to review psychosocial correlates of persistent insomnia in clinical samples and in the general population; (d) to model the interrelationships of psychosocial, medical, and sleep-related factors in successful adaptation; (e) to review the major findings from the recent DSM-IV field trials on sleep disorders; and (f) to review promising leads which might help to elucidate the neurobiological significance of sleep physiological abnormalities in major psychiatric disorders, such as depression, schizophrenia, and Alzheimer's dementia. We will conclude by highlighting promising directions for intervention research and recent methodological developments pertinent to such research. Readers wishing to pursue specific conceptual and methodologic issues in relation to sleep and affective disorders are referred to Kupfer and Reynolds (38) and to Benca et al. (3). With respect to the Benca et al. (3) meta-analysis of electroencephalographic (EEG) sleep measures in psychiatric disorders, we underscored the importance of several issues: (a) the use of multivariate approaches in the ascertainment of sensitivity and specificity; (b) the use of quantitative automated measured of rapid eye movement (REM) and delta activity; and (c) the need to consider EEG sleep measures from remitted depressed patients, off medication (39).


Epidemiological studies of insomnia have reported that between 10% and 20% of respondents characterize their sleep problems as severe or constant (4, 33, 41). Sleep maintenance insomnia is the most frequent problem, followed by difficulty falling asleep and then early morning awakening. Several important demographic trends also emerge: First, complaints of poor sleep or insomnia increase with age, apparently paralleling age-related changes in sleep-stage physiology (18, 73). In addition, younger insomniacs more frequently complain of difficulty falling asleep, whereas older patients often complain of middle and terminal insomnia. Second, women of all ages have more sleep complaints than men. Third, complaints of poor sleep are more common in lower socioeconomic groups. However, despite the large number of people with insomnia complaints, only one-third seek medical help for this problem (71).

Epidemiological studies have also documented a robust association between insomnia and psychiatric disorders (particularly depression and anxiety) across the life cycle. For example, in the Zurich study, where the course of insomnia was examined over 7 years in a cohort of young adults, the authors reported the following: (a) specific associations between recurrent brief insomnia and recurrent brief depression; (b) an association between continued insomnia and major depression; and (c) an association between any form of insomnia and anxiety disorders (69). The authors also observed that insomnia tends to recur, regardless of subtype. Similarly, reporting on a study of a mixed-age clinical sample of 954 psychiatric patients with major affective disorders, Fawcett et al. (17) observed six clinical features which were associated with suicide within 1 year. They included global insomnia, which is potentially a modifiable suicide risk factor. Similarly, the prevalence of persistent sleep disturbance is quite high in community resident samples, particularly among the elderly, where a 12–15% 6-month prevalence rate of persistent sleep disturbance was noted in the epidemiological catchment area (ECA) survey by Ford and Kamerow (19).

Persistent sleep disturbance was identified in the ECA data set as a highly significant risk factor for the subsequent development of major depression, as well as a major factor in patient and family decisions to seek services from the primary care sector. Hence, Ford and Kamerow (19) suggested that early intervention to treat sleep disturbance might protect patients from developing major depression.

In the setting of sleep, psychiatric morbidity, and quality of life, a longitudinal study of depressed mood and sleep disturbances in the elderly by Rodin et al. (55) examined the association between frequency of depressed mood and self-reports of four sleep problems over a 3-year period in a randomly selected sample of 198 community residents (mean age of 72). A robust association between frequency of depressed mood and severity of sleep symptoms was noted, after controlling for health status, gender, and age. Longitudinal fluctuation in sleep complaints (particularly sleep continuity disturbance and early morning awakening) covaried strongly with intensity of depressive symptoms. This observation was confirmed and extended by Kennedy et al. (34), who studied the persistence or remission of depressive symptoms in late life in a sample of 1885 adults aged 65 or older from the NIMH ECA study. The authors observed that changes in health, sleep disturbance, and added formal support services distinguished the persistently depressed and remitted groups.

Taken together, these studies across the life cycle suggest that sleep disturbance is a consequence of depressive symptoms, a significant correlate of help-seeking behavior, and a major risk factor for the subsequent development of syndromal major depression. Other studies have shown that sleep-related behaviors often precipitate the decision of families to institutionalize an elderly dementing relative (45, 48, 60). Any understanding of how to attenuate the sleep changes and disturbances of usual and pathological aging that would reduce psychiatric morbidity, burden to families, and the rate of institutionalization would be of enormous public health significance. Thus, we believe that the challenge to researchers and clinicians is to understand how the successful functioning of the aging circadian time-keeping system can be preserved in the face of multiple medical and psychosocial challenges (12).



As recently reviewed by Bliwise (5), there is now considerable evidence underscoring the general effects, either direct or indirect, of medical illness burden in poor sleep. While such difficulties often reflect presumably nonspecific effects of pain, impaired mobility, poor nutrition, and poor physical conditioning, specific clinical symptoms or disorders have also been shown to have negative effects upon sleep quality, including nocturia (16, 78), headache (13), gastrointestinal illness (40), bronchitis and asthma (40), cardiovascular symptoms (28), and Type II diabetes (28). Chronic pain from conditions like osteoarthritis also disrupts sleep in the elderly (74). In addition, elevated autonomic activity and a greater susceptibility to external arousal may be important predisposing factors to disturbed sleep, particularly in later life (46, 58, 76, 77).

Medical and neurological illness and cognitive impairment may be related not only to sleep, but to mortality as well. We have reported increases in sleep-disordered breathing, as well as deterioration in sleep continuity and REM sleep, in patients with Alzheimer's dementia (54). These changes correlate with 2-year mortality (27). Furthermore, Kripke et al. (36), Ancoli-Israel et al. (2), and Bliwise (5) have observed (a) high prevalence of sleep-disordered breathing with aging and (b) attendant increases in morbidity and mortality. Other data—for example, those of He et al. (26)—suggest, however, that sleep-disordered breathing may have a stronger relationship to mortality in midlife (where it is often more severe) than in late life (where it is typically subsyndromal).



As we have recently reviewed (14), sleep problems are often reported to occur in the context of, or to be exacerbated by, other life stressors. By far the most extensive body of work has focused on sleep disturbances in treated psychiatric patient populations documenting the significant comorbidity between sleep abnormalities and psychiatric disorder, particularly in the area of mood disorders (38). The descriptive, primarily cross-sectional investigations in this domain have detailed a range of abnormalities in sleep continuity and architecture and have also considered the clinical and sociodemographic correlates of these abnormalities across the lifespan, both with respect to subjective self-report measures (7) and with respect to objective polysomnographic indices (23, 27, 54). The age of subjects in particular appears to be an important moderator of the sleep changes that accompany psychiatric disorders (35), and late-age sleep deterioration in general has been found to be influenced by a variety of psychosocial factors, including gender (53), major life events such as bereavement (51), and ongoing strains such as those arising from chronic medical illness (5).

These clinical sleep studies, recognizing the multifactorial nature of life stresses and their correlates, have begun to expand the breadth of their psychosocial assessments in order to yield a more accurate and detailed picture of the social circumstances in which subjects' lives and their sleep are embedded. As Dew et al. (14) have pointed out, an important limitation of most existing studies is that the generalizability of findings from selected patient groups to nonpatient-community samples, the majority of whom do not seek treatment, is questionable. In fact, Vitiello et al. (67) reported that polysomnographic sleep is undisturbed in depressed individuals who have not sought health care.

Several studies of nonpatient, community-based samples have carefully examined (a) the roles of multiple psychosocial factors in altered sleep patterns and (b) the impact of sleep and sleep quality on subsequent quality of life. These reports provide evidence that diverse variables such as life events, psychological state, age, and gender are robustly correlated with reduced sleep quality (9, 55).

With few exceptions (9), however, studies employing nonpatient samples have relied on subjective reports of sleep problems, rather than objective polysomnographic data which is, by contrast, available in many clinical populations. Given this gap in the database, our group recently performed a study examining longitudinal data on psychosocial status and polysomnographic sleep collected annually from 57 healthy, community-residing elders aged 61–89 (14). Cluster analysis of variables reflecting sleep continuity and architecture at the baseline assessment identified three groups of elders: those whose sleep was either (a) superior to all remaining respondents across a variety of measures, (b) marred only by significantly reduced sleep efficiency relative to other respondents, or (c) poorer than all other respondents in multiple areas. Cross-validation procedures suggested that the three-group cluster solution was stable and replicable over persons and over time. Subsequent multivariate analyses indicated that recent life events, as well as psychosocial stability and support variables at baseline, distinguished the sleep-pattern groups. Moreover, sleep pattern group membership itself predicted subjects' subsequent sleep characteristics and psychosocial status at follow-up. Of special significance, the group of elderly subjects with poor sleep efficiency had more negative social circumstances at baseline, with an elevated number of major negative life events as well as less social stability and support (as reflected by reduced social rhythmicity, number of regular activities, and lower levels of perceived instrumental social support). This finding is consistent with the growing literature documenting the pathogenic role of life stress and the protective role of social stability and support with respect to mental health and well-being (for review focused on the elderly, see ref. 50).

These data led us to propose that specific psychosocial factors (e.g., major life events and social support) contribute to observed sleep variability. Furthermore, we hypothesize that EEG sleep disturbances will also themselves be likely to influence subsequent adaptation along physical, psychological, and psychosocial dimensions. These data clearly support the significance of considering a rich array of psychosocial variables in relation to sleep. Several points of intervention for improving sleep and waking life quality are also suggested: either by focusing on identified precursors of sleep disturbance (e.g., psychosocial stability and support) or by targeting elements of sleep which predict subsequent psychosocial changes.

Recent data on the sleep-disrupting effects of life events cast further light on these issues. Sleep, particularly dream sleep, may remain altered for prolonged periods of time after exceedingly negative or protracted stressors, as seen in the case of post-traumatic stress disorder (57). Indeed, a study of sleep in Holocaust survivors suggested that sleep disturbance with nightmares may persist for up to 50 years after extreme stressors (56).

We have suggested elsewhere that sleep may show greater alteration when the experience of a life event and major transition leads to significant and persistent changes in mood, as seen in major depression or post-traumatic stress syndrome. Thus, for example, depressed divorcing women (9) and depressed bereaved elders (51) are more likely to show shorter REM sleep latency or greater disruption in sleep continuity (with early morning awakening) than subjects who negotiate these life events without becoming depressed. Furthermore, bereaved subjects with depression of subsyndromal intensity complain of disturbed sleep but appear to have generally normal EEG sleep, with the possible exception of diminished slow-wave sleep generation in the first non-REM sleep period (44). These data suggest, therefore, that the development of major depression may mediate sleep changes following major life events or transitions. In the absence of pre-event sleep measures, however, this inference must be regarded as tentative.

These observations give rise to another important question, namely, whether there are any stable EEG sleep correlates of life stresses not confounded by concurrent major depression. This is an important issue theoretically and clinically because major life changes require adaptation to new roles and often to a new view of self. For example, to adapt to the loss of a spouse without becoming depressed could be regarded as an example of successful adaptation, in the sense of showing resilience, or the ability to bounce back from a major negative life event. The question therefore becomes, Does sleep remain entirely normal under such circumstances? Could sleep be an important correlate, or even a mechanism, of the resilience of successful adaptation in the context of major life transitions? If so, what aspects of sleep, if any, change during successful adaption, and what inference about the functions of sleep might one draw from the presence or absence of such changes?

REM sleep is a theoretically appealing candidate for providing a psychobiological correlate of successful adaptation to major changes in life, such as bereavement (50). Although we still do not know the functions of REM sleep, it has been suggested by Cartwright (9) that REM sleep may be a state of the mind brain during which affective information processing and affect discharge take place. If so, the burden of affective information processing during adaptation to a major life change may well lead to changes in the timing or intensity of REM sleep.

To pursue these issues, we recently performed a comparison of EEG sleep in nondepressed, spousally bereaved elders and in a healthy control group, in order to search for possible psychobiological correlates of bereavement not confounded by concurrent major depression (52). Bereaved and control subjects showed consistent differences over 2 years of follow-up in the phasic measures of REM sleep (increased in bereaved subjects), but were similar on all other EEG sleep measures over the 2-year period of observation. The bereaved subjects showed a small decline in the percentage of slow-wave sleep over the 2 years in which they were followed, but measures of sleep efficiency, REM latency, and delta sleep ratio were stable and did not differ from values seen in control subjects. Bereaved and control subjects were also similar on subjective sleep quality.

These data led us to conclude that during successful adaptation to the loss of a spouse, and in the absence of major depression, spousal bereavement is associated with elevation in the phasic measures of REM sleep, but does not appear to be associated with other physiological sleep changes typical of major depression when studied at 3–23 months after the spousal loss. This suggests that preservation of normal sleep following a major negative life event may be an important correlate, if not mechanism, of the resilience seen in successful adaptation to life events. Furthermore, the elevation in REM density may provide a psychobiological correlate of bereavement not confounded by concurrent major depression.


The perennial debate among sleep researchers about the relative importance of psychiatric and medical factors in insomnia led us to conceptualize a unified model of the multiple factors which challenge successful circadian time-keeping, including sleep, across the life cycle (50). This model provides a biopsychosocial framework in which to understand pathogenesis, assessment, and intervention for persistent insomnia. The model is based on the results of our own work, as well as on other data in the literature (reviewed above), and is shown in the Fig. 1.

The model explicitly shows important predictors of, as well as outcomes resulting from, the development of sleep and circadian function across the life cycle. Thus, changes in health and cognitive status and negative life events (particularly those associated with significant loss) are hypothesized to lead to decay in sleep and sleep quality. However, the model posits that much of the effect is not direct, but is instead mediated by two factors: changes in mood (negative shifts in affect balance) and worsening of sleep-disordered breathing. We also hypothesize that characteristics such as gender, stability of social rhythms, and social support serve as important moderators, helping to buffer the subject from the effects of medical burden, negative life events, and their mediators. It is also noteworthy that the model explicitly recognizes that sleep changes are also likely to influence subsequent adaptation to aging along physical and psychological dimensions. Thus, what appear as major predictors in the model may, over time, ultimately be influenced by the very outcomes that they have helped to produce. That is, the model also permits one to test empirically whether sleep itself indicates changes or stability in psychosocial variables over time.

Medical burden and increasing physical functional limitations, together with other negative life events and ongoing difficulties (e.g., financial strain, family member illnesses and death), represent chronic stresses in the model and are among the vulnerability factors most frequently addressed in previous research on the precursors of psychiatric symptoms and disorders (for review, see ref. 50). More specifically, chronic stressors that have received the most empirical scrutiny in mixed age samples are job stress, chronic financial strain, and chronic physical illness; all three are related to increased risk of psychiatric morbidity, particularly depression, anxiety, and alcohol misuse. Similarly, the primary protective factor examined in previous research is social support, including the dimensions of social network, tangible support, and perceived social support. However, we do not know whether or not support exerts its effects directly upon such outcomes as sleep, or primarily serves to buffer the impact of other stressors.

Our own work on sleep in bereavement provides evidence linking negative affect balance (i.e., depression) and sleep outcomes. Among recently bereaved elders, sleep is disturbed among widows and widowers who present with a full depressive syndrome. Sleep efficiency is diminished, REM latency is short, and early slow-wave sleep generation is reduced, compared to nondepressed recently bereaved subjects who are not depressed and compared to healthy elderly controls (51, 52). Similar findings linking stress, depression, and EEG sleep disturbances, including decreased delta sleep and shortened REM latency, were reported by Cartwright et al. (10) in a study of middle-aged divorcing spouses.

An understanding of the proposed biopsychosocial model is not complete without also considering sleep in pathological aging. Given the theoretical interest in the impact of medical burden and cognitive impairment on sleep, we have published the observation that Alzheimer's dementia, in contrast to major depression, is associated with deficits in the production of phasic activity during sleep, including rapid eye movements, sleep spindles, and K complexes (54). Alzheimer's dementia is also associated with increased rates of sleep-disordered breathing (27). Furthermore, in studies in patients with depressive pseudodementia, the sleep physiological response to one night of sleep deprivation is characterized by robust REM sleep rebound (6). Such a rebound is absent in patients with Alzheimer's dementia. The relevance of this observation to the current model and to its prediction that sleep outcomes influence subsequent adaptation is that sleep measures predict survival status at 2 years in patients with mixed clinical presentations of depression and cognitive impairment (27). Nonsurvivors show impaired capacity to generate REM sleep and to maintain respiratory control during sleep.


The proposed model certainly does not preclude the possibility that some illnesses directly exert their effects on sleep, rather than being mediated by other factors. This becomes clear in considerations of the neurocircuitry and neurotransmitters involved in sleep. Several neurotransmitter systems have been extensively studied in sleep, including serotonin, dopamine, acetacholine, and noradrenergic systems. While the data in this area are not easily summarized, two particular observations appear relevant to the current discussion. First, as reviewed by Meltzer and Lowy (42), "Decreased serotonergic activity is consistent with many of the changes in mood and somatic function observed in depressed patients, for example, decreased mood, insomnia, decreased REM latency, and disturbed circadian rhythms." Secondly, as reviewed by Janowsky and Risch (31), "Cholinergic agonists generally cause a shortening of REM latency and an increase in REM density, particularly in patients with an affective disorder episode or a family history of affective disorder." These observations have led to the hypothesis that changes in sleep in depressed patients may reflect neurotransmitter abnormalities, particularly an increased ratio of cholinergic to aminergic neurotransmission. Conversely, the decreased capacity for REM sleep generation in Alzheimer's dementia may reflect impairment, loss, and underactivity in cholinergic systems.

The availability of positron emission tomography (PET) and its utilization in psychiatric sleep research may lead to a more direct testing of these hypotheses, as suggested by the Institute of Medicine Panel on Basic Sleep Research (29). In this context there are several promising approaches which might elucidate the neurobiological significance of sleep physiological abnormalities, in major depression and schizophrenia. These approaches include studies of (a) the effects of family history of psychiatric disorder on sleep abnormalities, (b) the clinical correlates of EEG sleep abnormalities, and (c) functional imaging correlates of EEG sleep changes.

The ongoing work of Giles et al. (22, 23) illustrates the first approach, in which it is being asked whether EEG sleep is abnormal in persons at risk for major depression by virtue of family history. Data suggest that abnormal EEG sleep has shown an association within families, as well as a predictable relationship to lifetime development of depression (22) and an increased risk for new onset of depression (23).

Studies of clinical correlates of sleep abnormalities in schizophrenia have suggested that loss of slow-wave sleep (probably the most robust finding in schizophrenia) and evidence of cerebral atrophy on computerized tomography (CT) scan are strongly correlated with severity of negative symptoms (21, 65). Other studies in major depression have addressed EEG sleep correlates of long-term clinical response. For example, reduction in pre-treatment REM latency was found by Giles et al. (24) to be associated with higher risk for recurrence of major depression. Kupfer and co-workers (37, 38) reported that reductions in delta sleep ratio were associated with a significant increase in risk for recurrent major depression among midlife patients randomized to a maintenance psychotherapy condition.

Imaging correlates of EEG sleep changes in psychopathological states have now been conducted in patients with schizophrenia and major depression. These studies have led to several important observations: (a) increased cortical atrophy in relation to slow-wave sleep deficits in chronic schizophrenia (65) and (b) a finding of persistently increased limbic glucose metabolic rates in depressed patients who fail to show an antidepressant response to total sleep deprivation (75). In other words, total sleep deprivation appears to be associated with greater decreases in rates of glucose metabolism in limbic system among endogenously depressed patients who show an antidepressant response to this intervention.


By way of summarizing this discussion before proceeding to clinical issues, and attempting to tie together both neurobiological and psychosocial determinants of sleep changes, we are suggesting that successful adaptation across the life cycle is associated with preservation of sleep quality, ability to maintain daytime alertness, and physiological integrity of nocturnal EEG sleep (50). Failure to adapt is associated with loss of sleep continuity, alterations in the temporal distribution of delta wave activity, and either a relative increase of REM sleep (mood disorders) or a decrease in REM sleep (neurodegenerative disorders). As noted above, the increase in REM sleep which accompanies depression, whether endogenous or in the wake of negative life events, is also correlated with an attendant decrease in positive affect balance and diminished stability of social rhythms in the case of bereavement. By contrast, diminished capacity for REM sleep generation seems to accompany neurodegenerative disorders such as Alzheimer's dementia and is a correlate not only of cognitive impairment, but also of early mortality. The model we propose attempts to account for data at different points along the continuum of aging, from successful to pathological. Clearly, further longitudinal studies will be required to validate the hypothesized interrelationships among sleep, aging, neuromedical, and psychosocial variables.

In the next section of this chapter, we shall shift to clinical considerations, such as issues of diagnostic classification, the utility of polysomnography in the evaluation of chronic insomnia, the impact of diagnosis and interviewer experience on treatment recommendations, and promising areas for intervention research.



Because of long recognized links between certain types of sleep disorders (particularly insomnias) and psychiatric symptoms, psychiatry has played an integral role in the development of sleep disorders medicine over the past quarter century. Furthermore, given the clear association between sleep and psychiatric disorders, psychiatric clinicians and researchers need a comprehensible, reliable, and valid sleep disorders nosology. This need provided the impetus for the recently completed APA/NIMH/DSM-IV field trial on the diagnostic reliability of sleep disorders.

Several diagnostic classifications have been developed for sleep disorders in the past 15 years, including the Diagnostic Classification of Sleep and Arousal Disorders, 1st edition [DCSAD; American Sleep Disorders Association (62)], the Diagnostic and Statistical Manual, 3rd edition, revised [DSM-IIIR; American Psychiatric Association (1)], and the International Classification of Sleep Disorders [ICSD, 1991 (15)]. These systems differ with respect to (a) broad versus detailed diagnoses, (b) organizing categories by presenting symptom or presumed pathophysiology, and (c) intended users (general clinicians versus sleep specialists). Only DSM-IIIR and ICSD include specific diagnostic criteria, and only ICSD includes polysomnographic features among those criteria (although polysomnographic findings are not required to make a diagnosis). Recently, Schramm et al. (61) in Germany have completed a study showing excellent rates of interrater agreement (kappas in excess of .80) for DSM-IIIR sleep disorders, when sleep disorder specialists used a structured clinical interview to make diagnoses.

The sleep disorder section in the Diagnostic and Statistical Manual, 4th edition (DSM-IV) combines elements of previous diagnostic systems and takes into account controversies surrounding sleep disorders diagnosis, such as broad versus narrow diagnoses, the role of EEG sleep studies in diagnosis, and the intended users (see Table 1). DSM-IV includes a greater number of more specific categories than does DSM-IIIR (23 versus 15, including subtypes), but far fewer than the 84 total disorders enumerated in the ICSD. Like ICSD, DSM-IV includes major categories based on presumed pathophysiology, rather than on symptoms alone. Another proposed nosology, the International Classification of Diseases, 10th edition (ICD-10), is similar to DSM-IV in the number of specific diagnoses (17 total), but has only two general categories compared to four for DSM-IV. Neither DSM-IV nor ICD-10 include specific polysomnographic criteria. (An outline of DSM-IV and ICSD is presented in Table 1).

The DSM-IV multisite field trial, sponsored by the American Psychiatric Association and the National Institute of Mental Health, was undertaken to determine rates of interrater agreement for sleep disorder diagnoses using the proposed classification systems, DSM-IV and ICD-10. The field trial focused on patients with insomnia complaints, because these are most relevant to psychiatric practice (Buysse et al., in press). Five sites interviewed a total of 257 patients (216 referred for insomnia complaints, and 41 control patients). One sleep specialist and one general clinician interviewed each patient, using a nonstructured clinical interview, and assigned DSM-IV and ICD-10 diagnoses.

We found that "insomnia due to another mental disorder" was the most frequent DSM-IV diagnosis, occurring in 76% of cases, as either a primary (43%) or secondary (33%) diagnosis. "Primary insomnia" was the next most frequent diagnosis, occurring in 48% of cases as either the primary (21%) or secondary (27%) diagnosis. "Insomnia due to emotional causes" was the most frequent ICD-10 diagnosis, occurring in 62% of cases. Interrater kappa value for multiple DSM-IV diagnoses was 0.47, with considerable range for individual diagnoses from 0 to 0.70. Individual sites varied considerably in the distribution of diagnoses, as well as in rates interrater agreement. Interviewers indicated greater confidence and better fit of diagnoses, as well as greater ease of use, for DSM-IV as compared to ICD-10.

The distribution of insomnia diagnoses in the DSM-IV field trial was similar, with some exceptions, to that reported by Coleman et al. (11), who used the Diagnostic Classification of Sleep and Arousal Disorders (ASDA) (62). These results confirm the importance of psychiatric and behavioral factors in clinicians' assessments of insomnia patients, and they suggest that ICSD and DSM-IV sleep disorder diagnoses have similar patterns of use by experienced clinicians. The results also suggested that rates of agreement were higher for more specific diagnostic categories and lower for less-well-defined diagnoses. Furthermore, site-related differences appeared to occur not only because of different patient populations but also because of different application of diagnostic criteria. In this context, the differentiation of primary from psychiatric forms of insomnia was frequently a source of disagreement between raters.

Although in both the DSM-IV field trial and the Coleman et al. (11) multisite trial psychiatric insomnia was the most prevalent, followed by psychophysiological (primary) insomnia, the DSM-IV field trial actually found a somewhat higher percentage of psychiatric insomnia (40% versus 35%) and a slightly lower percentage of psychophysiological (primary) insomnia (12.5% versus 15.3%). Other differences included in the DSM-IV field trial were lower percentages of substance-related insomnia (3.1% versus 12.4%) and periodic limb movements/restless legs (1.2% versus 12.2%), but a higher frequency of delayed sleep phase syndrome (7.0% versus approximately 3.6%) and a similar frequency of sleep apnea (5.4% versus 6.2%). In comparing these studies, it is important to note that Coleman et al. (14) collected diagnoses after polysomnography, while diagnoses in the DSM-IV field trial were assigned on purely clinical grounds after one initial interview.

An important conceptual difference between DSM-IV and ICSD consists of the greater subtyping of nonpsychiatric, nonmedical forms of insomnia in ICSD. Patients diagnosed with DSM-IV "primary insomnia" actually fall into three major ICSD categories: "psychophysiological insomnia," "inadequate sleep hygiene," and "idiopathic (childhood-onset) insomnia." This type of difference relates to the conceptualization and presumed etiology of insomnia problems. Thus, in DSM-IV, "primary insomnia" is essentially atheoretical with regard to causation; "psychophysiological insomnia" in ICSD invokes conditioning and physiological factors; "inadequate sleep hygiene" attributes causation to voluntary behaviors; and "idiopathic insomnia" assumes a yet-to-be identified genetic or biological component. Whether the broader DSM-IV (and ICD-10) categories or the more specific ICSD categories have greater validity remains to be determined. Indications of validity may include differences in polysomnographic measure, longitudinal course, treatment response, and familial patterns. We have suggested elsewhere that such indicators of validity have not yet been demonstrated (53).

Although the overall rate of agreement for DSM-IV diagnoses was moderately good, it was lower than that usually reported for psychiatric disorders. We speculate that several factors probably contributed to the lower rate of agreement: (a) the absence of a structured interview; (b) the use of interviewers without specialized training in sleep disorders; and (c) relatively low base rates of certain disorders, which had the effect of lowering kappa values. These results provide a conservative estimate of diagnostic reliability—higher than that which may be seen in routine clinical practice where even more heterogeneous interviewers and lower rates of insomnias may occur, but lower than that which could be obtained using structured interviews and formal training (72). Finally, the higher kappa values seen for DSM-IV compared to ICD-10 diagnoses suggest that not only the total number of diagnostic categories, but also their familiarity and sensibility to interviewers, may relate to interrater agreement.



The findings of the DSM-IV field trial, together with the model proposed in this chapter, lend themselves to the formation of a practical algorithm for the diagnosis of DSM-IV sleep disorders. The intent of this algorithm, created by Nancy Vettorello and Harold Alan Pincus of the APA Office of Research (personal communication to Charles Reynolds and David Kupfer), is to translate into usable clinical form the results of the field trial and the findings reviewed from the literature and incorporated into our model of the medical and psychosocial correlates of sleep disturbance.

The clinician who is confronted by a complaint of sleep disturbance on the part of the patient or a family member must define the nature of the complaint, whether primarily of insomnia, excessive daytime sleepiness, disturbed mentation or behavior during sleep, or difficulties in the circadian placement of sleep. The first step in the diagnostic algorithm is to consider the patient's general medical condition, in order to determine whether the patient's complaint represents a secondary sleep disorder due to a general medical condition. Furthermore, if the patient is taking medication or using a substance, the clinician should consider the diagnostic possibility of a substance-induced sleep disorder (step 2 of the DSM-IV sleep disorders algorithm).

If sleep disturbance is suspected to be related to another mental disorder, the clinician should consider diagnoses such as major depression, anxiety disorders, or post-traumatic stress disorder as the etiology of the sleep complaint (step 3). If the sleep problem, particularly one of excessive daytime sleepiness, occurs in the context of a history of snoring or with obesity, the clinician should consider a breathing-related sleep disorder (step 4). If the patient frequently crosses time zones or is involved in shift work, or has a primary problem of sleep timing (rather than disturbance), a circadian rhythm sleep disorder or a sleep–wake schedule disorder should be considered (step 5).

If the patient's symptoms are predominantly behavioral or mental events during sleep (e.g., abrupt awakening, frightening dreams, or walking about while sleeping), the clinician should consider a diagnosis of nightmare disorder, sleep terror disorder, sleep walking disorder, or another type of parasomnia (step 6).

If the primary complaint is persistent insomnia and/or difficulty initiating or maintaining sleep, a diagnosis of primary insomnia is suggested if (a) symptoms have persisted for more than a month and (b) symptoms do not appear to be associated with another mental or physical disorder (step 7). Similarly, if the primary complaint is excessive sleepiness, the clinician should consider the differential diagnosis of a breathing-related sleep disorder, narcolepsy, or primary hypersomnia, in the absence of a mental or physical disorder which could explain the complaint (step 8).

Finally, if the criteria for previously described specific disorders are not met, a "default" diagnosis of dysomnia not otherwise specified should initiate a more complete evaluation (step 9).

For further reference, Table 1 summarizes the DSM-IV sleep–wake disorders nosology and the International Classification of Sleep Disorders (ICSD).


The differential diagnosis of chronic insomnia inevitably leads to a review of the current controversies surrounding the use of polysomnography in the evaluation of insomnia, the subject of a recent task force report by the American Sleep Disorders Association (49). Two of the authors (Reynolds and Buysse) participated in the writing of this task force report. The central issue addressed in this report is whether polysomnographic findings are helpful clinically—that is, do they elucidate pathophysiology, establish a diagnosis, or guide treatment recommendations? The task force report concluded that "sleep studies are more likely to be helpful if there is a specific suspicion of an intrinsic sleep pathology (such as sleep apnea or periodic limb movements), or if the sleep complaints are not adequately explained by the type or degree of medical illness and medications. Examples would include a suspicion of periodic leg movements in a patient with chronic renal disease; a suspicion of apnea in a patient with neuromuscular or chronic pulmonary disease; or a suspicion of a seizure disorder in a patient who has suffered head trauma or stroke." Furthermore, as reviewed by Bliwise (5), older patients more frequently demonstrate unsuspected periodic leg movements and apnea. The incidence of secondary sleep disorders due to medical illness increases with age, suggesting that polysomnographic findings are more likely to be helpful in older patients. With respect to the goal of separating subtypes of insomnia, the task force report indicated that a more comprehensive assessment than is typically performed might be useful. Areas of potential interest include assessment of the circadian system (e.g., systematic temperature assessment) and use of computerized EEG analysis (e.g., period analysis and power spectral analysis).

The ASDA task force authors suggested that "far more important than polysomnography is a careful and thorough clinical evaluation of the patient with an insomnia complaint, and perhaps a therapeutic trial based upon the most probable cause(s), before polysomnography, in most cases." At the same time, however, the authors acknowledged that "there is no general agreement as to what constitutes an adequate therapeutic trial." Whatever approaches may be taken, we suggest that follow-up assessment is important in order to determine the stability of clinical response. With respect to the conduct of clinical trials, we also advocate the use of specific diagnostic criteria, such as those tested in the field trial, in order to constitute more homogeneous study groups. Finally, we advocate the use of both subjective and objective outcome measures in clinical trials of interventions in insomnia.


The DSM-IV field trial was also designed to permit an examination of (a) whether different insomnia diagnoses are associated with different treatment recommendations and (b) whether sleep specialists and nonspecialists advise different treatments. The trial found that treatment and polysomnography recommendations differed significantly for different DSM-IV diagnoses (Buysse et al., unpublished observations). Sleep specialists recommended several behavioral strategies—including self-monitoring, sleep hygiene techniques, and stimulus control techniques—more highly than did nonspecialists. However, nonspecialists more strongly recommended relaxation training and medications for medical or neurological conditions. Nonspecialists also recommended polysomnography more strongly than did specialists.

Most treatment recommendations differed significantly for different diagnoses. For instance, position training, continuous positive airway pressure, and oxygen were all recommended more strongly for "breathing-related sleep disorder" than for the other diagnoses; psychiatric medications and interventions were more strongly recommended for "insomnia related to another mental disorder"; and behavioral intervention were more strongly recommended for "primary insomnia." Polysomnography was recommended for "breathing-related sleep disorders" more strongly than for other diagnoses.

Significant site-related differences also emerged in the treatment recommendations, even for the same diagnoses. Such site-related differences reflect the perennial debate in the sleep-disorders and research community about (a) the importance of nonpsychiatric factors in insomnia and (b) the proper method of subtyping insomnia diagnoses. On the one hand, numerous studies have shown elevations in MMPI scores and other measures of psychopathology in chronic insomnia patients, including categorical psychiatric diagnoses (30, 32, 64). These findings can lead to the conclusion that chronic insomnia is almost always associated with significant psychopathology, and that psychiatric treatment is indicated. On the other hand, other studies have shown that chronic insomnia not uncommonly results from either (a) more specific intrinsic disorders of sleep (e.g., periodic limb movements, sleep apnea), (b) concurrent medical or neurological illness, or (c) behavioral and conditioning factors (11, 30). Such studies emphasize diversity in etiology—and presumably in the treatment—of chronic insomnia. Thus, site-related differences in the DSM-IV field trial suggest that some controversy remains in the conceptualization of sleep disorders, their evaluation, and their treatment. The array of contemporary treatment options for chronic insomnia is shown in Table 2, where they are grouped according to psychiatric, behavioral, and neuromedical interventions.



In concert with the 1990 NIH Consensus Development Conference on Sleep Disorders in Late Life, the authors believe that chronic primary insomnia, particularly sleep-maintenance insomnia, represents a significant public-health problem and an important opportunity for intervention research. The authors also suggest that, to be truly successful, intervention must be shown to be efficacious over long periods of time (because primary insomnia tends to be chronic and recurring). Based upon the work of Spielman et al. (63) and Friedman et al. (20), we suggest that therapy utilizing sleep-restricting techniques holds particular promise for short- and long-term efficacy in the management of primary, sleep maintenance insomnia, particularly in late life. Another promising lead, though less well researched to date, is the use of bright-light exposure in the evening (8). Properly timed bright-light exposure may help reduce sleep-maintenance difficulties and improve daytime alertness, possibly via internal phase realignments of circadian temperature and sleep–wake rhythms. Finally, based upon the work of Vitiello et al. (68), enhancing aerobic fitness may be associated with improved sleep quality.

Finally, although we endorse the need for further controlled clinical trials to establish the short- and long-term efficacy of behavioral, psychotherapeutic, and naturalistic interventions for primary insomnia, we suggest that it would be premature to exclude a role for maintenance medication. Because significant progress has been made in the methodology of maintenance therapy trials, we endorse the application of such methodology to long-term intervention studies of chronic insomnia. We suggest that antidepressant medication or intermittent benzodiazepines could justifiably be investigated for chronic efficacy in sleep maintenance insomnia.

In conclusion, although the interventions which the authors regard as promising candidates for treatment research in chronic insomnia are quite diverse, such diversity attests to the heterogeneity of chronic insomnia. Conceptually, it is plausible to suggest that chronic primary insomnia is the behavioral expression of many different underlying factors. In some patients, diminished sleep drive will predominate; in others, misalignment of circadian rhythms will be etiologically important; and in still others, mood disorders, often of subsyndromal intensity, will lead to insomnia. Such heterogeneity provides, we suggest, a compelling theoretical rationale for different interventions and, possibly, a framework for understanding differential treatment response. It also underscores the need for the use of a more rigorous, structured approach to diagnosis for defining study groups. The possibility for more rigorous sample definition has been advanced, we feel, through the DSM-IV insomnia field study.



This work was supported, in part, by NIMH grants MH30915, MH00295, MH37869, MH48891, MH47200.


published 2000