- Email: [email protected]
Assessing insomnia severity in depression: comparison of depression rating scales and sleep diaries
JOURNAL OF PSYCHIATRIC RESEARCH
Journal of Psychiatric Research 39 (2005) 481–488
www.elsevier.com/locate/jpsychires
Assessing insomnia severity in depression: comparison of depression rating scales and sleep diaries Rachel Manber a,*, Christine Blasey a, Bruce Arnow a, John C. Markowitz b, Michael E. Thase c, A. John Rush d, Frank Dowling e, James Koscis b, Madhukar Trivedi d, Martin B. Keller f a
c
Department of Psychiatry and Behavioral Sciences, Stanford University, 401 Quarry Road, Suite 3301, Stanford, CA 94305 650, USA b Department of Psychiatry, Cornel University Medical College, New York, NY, USA Department of Psychiatry, University of Pittsburgh School of medicine, Western Psychiatric Institute and Clinics, Pittsburgh, PA, USA d Department of Psychiatry, University of Texas Southwestern Medical Center, Dallas, TX, USA e Department of Psychology, State University of New York at Stony Brook, Stony Brook, NY, USA f Department of Psychiatry and Behavioral Sciences, Butler Hospital, Brown University, Providence, RI, USA Received 21 July 2004; received in revised form 17 November 2004; accepted 3 December 2004
Abstract Depression and sleep researchers typically assess insomnia severity differently. Whereas depression researchers usually assess insomnia with items on depression symptom inventories, sleep researchers usually assess the subjective experience of insomnia with sleep diaries. The present manuscript utilizes baseline data from 397 participants in a large multi-site chronic depression study to assess agreement between these two methodologies. The results indicate that the early, middle, and late insomnia items of the Hamilton Rating Scale for Depression (HRSD24) and the Inventory of Depression Symptoms – Self Report (IDS-SR30) are highly correlated with the weekly mean values of time to sleep onset, time awake after sleep onset, and time awake prior to the planned wake-up obtained from prospective sleep diaries. Results also reveal significant correspondence between the weekly-mean of daily sleep efficiency, an accepted measure of sleep continuity (the ratio between reported time asleep and time in bed), and the insomnia scale scores of the HRSD24 and the IDS-SR30 (the mean score on the three insomnia items of each depression measure). Unit increments in HRSD24 scores for early, middle and late insomnia were associated with significant increases in unwanted minutes awake for corresponding periods on sleep diaries. Similar relationships were found for early insomnia on the IDS-SR30 but not for middle and late insomnia. Overall, with few exceptions, findings revealed substantial agreement between the HRSD24, IDS-SR30 and prospective sleep diary data. The study supports the validity of the sleep items and sleep subscales of the HRSD24 and the IDS-SR30 as global measures of insomnia severity in depression. Conventional sleep assessment procedures can complement depression scales by providing additional information about specific aspects of sleep in depression. Ó 2005 Elsevier Ltd. All rights reserved. Keywords: Early insomnia; Middle insomnia; Late insomnia; Depression; Assessment
Sleep disturbance is a common and key depressive symptom that affects both the course of illness and treat*
Corresponding author. Tel.: +1 650 724 2377; fax: +1 650 725 8910. E-mail address: [email protected] (R. Manber). 0022-3956/$ - see front matter Ó 2005 Elsevier Ltd. All rights reserved. doi:10.1016/j.jpsychires.2004.12.003
ment response. Patients with abnormal sleep profiles have demonstrated significantly worse outcome to psychotherapy than depressed patients with normal sleep profiles in terms of symptom severity, response, remission, and attrition (Thase et al., 1996, 1997). In addition, subjective complaints of insomnia appear to affect the
482
R. Manber et al. / Journal of Psychiatric Research 39 (2005) 481–488
stability of response to depression treatment with a combination of psychotherapy and pharmacotherapy and the time-course to relapse (Dew et al., 1997). Disturbed sleep is also the most common residual symptom following the acute phase of fluoxetine therapy of major depression, present in approximately 44% of treatment completers (Nierenberg et al., 1999). The presence of insomnia at the end of treatment predicts earlier and more frequent relapse (Reynolds et al., 1997; Van Londen et al., 1998). Despite the significance of insomnia, depression researchers typically do not assess insomnia as comprehensively as sleep researchers. In most insomnia studies, sleep disturbance is measured with a prospective selfreport method, such as a sleep diary, and/or with an objective measure, such as an all night polysomnographic study or a sensitive activity monitoring device from which data on sleep and wakefulness can be extracted. On the other hand, depression researchers abstract an index of insomnia severity from items on standardized questionnaires or structured clinical interviews used to assess depression severity. For example, some investigators use the average of the insomnia items of the Hamilton Rating Scale for Depression, (Hamilton, 1960) or the Inventory of Depression Symptoms – Self Report (IDS-SR, Rush et al., 1986, 1996) to track change in insomnia severity during treatment for depression (Maes et al., 1993; McCall et al., 2000; Thase et al., 2002). While the HRSD24 and IDS-SR30 have good internal consistency and test–retest reliability (RamosBrieva and Cordero-Villafila, 1988; Rush et al., 1996), the validity of their individual insomnia items and insomnia subscale have not been investigated. The present study is the first to evaluate agreement between retrospective measures of insomnia derived from items on the aforementioned depression inventories and a week of prospective sleep diary data. The sleep diary was selected as the primary sleep measure because it is inexpensive, relatively easy to administer in a large, multi-site trial, and it provides prospective data on participantsÕ perceived sleep. IndividualsÕ perceptions of their sleep provide clinically meaningful data, as it is the subjective perception of sleep that guides help-seeking behaviors. We previously reported that both the insomnia scales of depression inventories (the HRSD24 the IDS-SR30) and measures derived from a sleep diary were sensitive to differential changes in perceived sleep difficulties associated with three treatment modalities, medication, psychotherapy, and their combination (Manber et al., 2003; Thase and Jindal, 2002). In the present manuscript we present findings on the relationship between the weekly mean values of time to sleep onset, time awake after sleep onset, and time awake prior to planned rise time, and the corresponding insomnia items of the IDS-SR30 and the HRSD24 (early, middle, and late insomnia). This
report also examines the relationship between sleep efficiency, a global measure of disturbed sleep that is commonly derived from the sleep diary (defined below), and global measures of insomnia severity derived from the two depression measures.
1. Method 1.1. Design Sleep data were collected as part of a large 12-site, randomized, chronic depression study that compared the efficacy of Cognitive Behavioral Analysis System of Psychotherapy (CBASP), a structured, short-term psychotherapy developed specifically for chronic depression (McCullough, 2000), the antidepressant medication nefazodone, and the combination of CBASP and nefazodone for patients with chronic depression (Keller et al., 2000). For a comprehensive description of the methods employed in study, we refer the reader to Keller et al. (2000). This report is based on data collected at baseline from participants who provided sleep data on the three HRSD24 and IDS-SR30 sleep items (i.e., early, middle, and late insomnia) and on the three corresponding sleep diary items (i.e., time to fall asleep time awake, in the middle of the night, and time awake before planned). 1.2. Participants The current study sample, the Sleep Diary Sample consisted of 397 participants who completed sleep diary data prior to randomization. The sample constitutes 58% of the randomized sample of the parent study (n = 681) and had virtually identical demographic characteristics, shown in Table 1 (Keller et al., 2000) 1. Participants were adult outpatients (66.8% female) between the ages of 18 and 75 (mean age 44.2 years, SD = 10.9 years). Eligible participants met Diagnostic and Statistical Manual of Mental Disorders, fourth edition (American Psychiatric Association, 1994) criteria for a current chronic major depressive episode that was either at least 2 years of duration or was superimposed on a preexisting dysthymic disorder, or was the most recent major depressive episode (at least 2 years duration) in a prior recurrent MDD with incomplete inter-episode recovery. Eligibility for the larger study, and hence for the current sample, required an age between 18 and 75 1 There were, however, statistically significant differences in HRSD middle and late insomnia items with those who completed the sleep diary (the present sample) having higher scores than those who did not complete the sleep diary and were therefore not included in the study (1.2 ± 0.8 versus 1.4 ± 0.8 for middle insomnia and 0.9 ± 0.9 versus 1.0 ± 0.9 for late insomnia p < .05).
R. Manber et al. / Journal of Psychiatric Research 39 (2005) 481–488
483
Table 1 Demographic characteristics of current (n = 397) and original sample (n = 681)
Age (years) Female gender (%) White race (%) Marital status (%) Married/cohabitating Single Widowed Divorced/separated Depression diagnosis (%) Chronic major depression MDD with dysthymic disorder Recurrent depression, with incomplete Inter-episode recovery HRSD24 score ± SD HRSD early insomnia score ± SD HRSD middle insomnia score ± SD HRSD late insomnia score ± SD
Sleep diary sample N = 397
RCT sample N = 681
44.2 ± 10.9 66.8 91.7
43.0 ± 10.7 65.3 90.5
47.3 24.4 2.3 25.9
42.7 27.2 2.1 28.0
33.2 45.1 21.7
35.1 42.3 22.6
26.9 ± 5.0 0.92 ± 0.89 1.41 ± 0.80 1.02 ± 0.89
26.9 ± 5.0 0.92 ± 0.89 1.3 ± 0.82 0.95 ± 0.89
and a score of at least 20 on the 24-item HRSD. Participants were excluded for the following reasons: a history of psychotic, bipolar or obsessive compulsive disorder; eating disorders within the past year; substance abuse or dependence within the past six months; a high risk for suicide; antisocial, schizotypal, or severe borderline (i.e., high risk for hospitalization or significant self-mutilation) personality disorder; dementia; seizure disorder; poorly controlled or serious medical disorders; and a history of failing adequate trials of two different types of antidepressant medications or two different courses of empirically supported psychotherapy for depression within the past 3 years. Diagnoses were derived using a modified version of the Structured Clinical Interview for the DSM-IV-R (First et al., 1995) for Axis I disorders and an abbreviated version of the SCID-II (First et al., 1997) for Axis II disorders (limited to antisocial, borderline, schizotypal, avoidant, dependent, and obsessive–compulsive personality disorders). All participants had a physical examination, routine laboratory tests, and urine toxicology screen. All patients provided written informed consent, approved by the relevant institutionÕs human subjects committee before entering the study. 1.3. Measures 1.3.1. Sleep diary data Sleep diaries were completed daily (prospectively) for one week before treatment began and before the baseline assessments of depression severity. Diary entries included bedtime, lights out time, actual wake-up time, planned wake-up time, last time out of bed, latency to sleep onset (minutes), and minutes of wakefulness after sleep onset (WASO). The following variables were computed from the diary entries: (1) time in bed (TIB), defined as the time
elapsed from bedtime until the last time out of bed; (2) total sleep time (TST), defined as the time elapsed from bed time until wake up time minus latency to sleep onset and WASO; (3) sleep efficiency (SE), defined as the ratio of TST to TIB; and (4) early morning awakening (EMA), defined as the difference between actual wake up time and planned wake up time; a negative value for EMA reflects an actual wake up time that is earlier than the planned wake up time. 1.3.2. HRSD Insomnia items The 24-item HRSD, an observer-rated measure of depression severity, includes three items that assess disturbed sleep. Although factor analytic studies of the HRSD24 have yielded inconsistent results, the three insomnia items of the HRSD24 (early, middle, and late insomnia), each rated by the interviewer on a scale ranging from 0 (absent) to 2 (severe), frequently load on the same HRSD24 factor (Fleck et al., 1995; Mowbray, 1972; Ramos-Brieva and Cordero-Villafila, 1988). The insomnia scale score is computed as the average of the three insomnia items, and therefore has a range between 0 and 2. The HRSD insomnia scale score can detect improvement in insomnia associated with effective treatment of depression (Thase et al., 2002) and has been shown to be sensitive to the differential impact of psychotherapy and antidepressant medications on sleep (Thase et al., 2002). In the present study, HRSD raters received external certification of competence and were unaware of patientsÕ treatment conditions. 1.3.3. IDS-SR insomnia items The 30-item IDS-SR is a self-rating scale of depression severity and includes three insomnia items (early, middle, and late insomnia), each rated on a 0–3 scale. The early insomnia item is anchored as follows: 0 = I never take longer than 30 min to fall asleep; 1 = I take
484
R. Manber et al. / Journal of Psychiatric Research 39 (2005) 481–488
Table 2 Spearman correlations between HRSD24 Insomnia items and sleep diary data (n = 397) and between IDS-SR30 insomnia items and sleep diary data (n = 384) Sleep diary
Early insomnia
Middle insomnia
Late insomnia
HRSD24 insomnia items Minutes to fall asleep Minutes awake after sleep onset Minutes awake before planned
+0.63 +0.16 0.01
+0.17 +0.49 0.26
+0.11 +0.26 0.49
IDS-SR30 insomnia items Minutes to fall asleep Minutes awake after sleep onset Minutes awake before planned
+0.70 +0.26 +0.02
+0.21 +0.60 0.29
+0.13 +0.31 0.46
at least 30 min to fall asleep, less than half the time; 2 = I take at least 30 min to fall asleep, more than half the time; and 3 = I take more than 60 min to fall asleep, more than half the time. The middle insomnia item is anchored as follows: 0 = I do not wake up at night; 1 = I have a restless, light sleep with a few brief awakenings each night; 2 = I wake up at least once a night, but I go back to sleep easily; 3 = I awaken more than once a night and stay awake for 20 min or more, more than half the time. The late insomnia item is anchored as follows: 0 = Most of the time, I awaken no more than 30 min before I need to get up; 1 = More than half the time, I awaken more than 30 min before I need to get up; 2 = I almost always awaken at least 1 h or so before I need to, but I go back to sleep eventually; 3 = I awaken at least 1 h before I need to, and cannot go back to sleep. The insomnia scale score is computed as the average of the three insomnia items, and therefore has a range be-
tween 0 and 3. The IDS-SR30 insomnia scale score can detect improvement in insomnia associated with effective treatment of depression and it has demonstrated sensitivity to the differential impact of psychotherapy and antidepressant medications on sleep (Thase et al., 2002).
2. Results 2.1. Concordance of sleep diary data and HRSD24 Moderately statistically significant Spearman correlations, shown on the diagonal of Table 2, were observed between HRSD24 and corresponding sleep diary responses. Off diagonal entries in Table 2 are correlations between HRSD24 and non-corresponding sleep diary items. Comparisons of off-diagonal entries in Table 2
Table 3 Mean ± SD and (Median) time (minutes) from the sleep diary for each level of the corresponding item on the HRSD24 (n = 397) and the IDS-SR30 (n = 384)
HRSD24 0 1 2 Paired comparisons IDS-SR30 0 1 2 3 Paired comparisons a
Earlya
Middlea
Latea
Time to sleep onset
Time awake after sleep onset
Time awake before planned
21.6 ± 19.8 (16.2) 39.0 ± 23.4 (30.0) 67.2 ± 43.8 (58.8) 0<1<2
16.3 ± 24.1 (7.1) 21.2 ± 23.7 (13.5) 47.0 ± 39.9 (36.9) 0 < 1*, 1 < 2
17.4 ± 78.0 ( 10.8) 17.4 ± 78.0 (16.2) 66.0 ± 90.0 (54.6) 0>1>2
17.8 + 19.7 (12.5) 31.6 + 21.6 (25.0) 49.4 + 29.3 (38.9) 84.7 + 45.8 (76.4) 0<1<2<3
3.5 + 4.5 (1.0) 26.4 + 25.6 (16.4) 17.7 + 16.2 (13.3) 54.8 + 32.2 (51.4) 0 < 1, 1 > 2*, 2 < 3
11.9 + 74.0 (6.4) 40.3 + 59.3 ( 32.5) 33.3 + 110.4 ( 28.2) 85.7 + 86.3 ( 73.6) 0 > 1, 1 = 2, 2 > 3
Kruskal–Wallis H test indicated significant differences in times by HRSD24 and IDS-SR30 rating for each insomnia item, all p values < 0.0001. Significant follow-up comparisons, using Mann–Whitney, are shown in the last row of each table, all p values, except the one marked with * < 0.001 (*p < 0.05). For late insomnia, a negative number represents a wakeup time that is earlier than planned.
R. Manber et al. / Journal of Psychiatric Research 39 (2005) 481–488
2.2. Concordance of sleep diary data and IDS-SR30 IDS-SR30 data were available for the majority of participants in the sample (n = 384, 97%). Shown on the diagonal of Table 2, Spearman correlations between sleep diary data and their corresponding IDS-SR30 items were +0.70 (early), +0.60 (middle), and 0.46 (late). Correlations off the diagonal were lower; effect sizes in terms of r2 were half of those observed on the diagonal, indicating substantial item discrimination. The mean and median times in minutes from sleep diary items for each level of the corresponding IDSSR30 item are depicted in Table 3. Kruskal–Wallis H
(a) 2.5
Mean of 3 HAMD Sleep Items
with those on the diagonal indicate substantial item discrimination, particularly for early insomnia. The HRSD24 early insomnia item was significantly and positively correlated with its corresponding item in the sleep diary, minutes to fall asleep (q = 0.63, df = 395, p < 0.0001). The HRSD24 middle insomnia item was significantly and positively correlated with diary recordings of minutes awake after sleep onset (q = 0.49, df = 395, p < 0.0001). The HRSD24 late insomnia item was inversely related to diary recordings of minutes awake before planned (q = 0.49, df = 365, p < 0.0001). The mean and median times in minutes from sleep diary items for each level of the corresponding HRSD24 item are depicted in Table 3. Kruskal–Wallis H tests followed by Mann–Whitney comparisons confirm that the means in number of minutes to fall asleep, minutes awake after sleep onset, and minutes awake before planned were statistically significantly different among the three levels of each of the three corresponding insomnia items on the HRSD24 (p < 0.001). For early and late insomnia increased scores on the HRSD24 were associated with significantly more minutes to fall asleep and with significantly more minutes awake in the morning before planned, respectively (all pÕs < 0.001). Increased scores on the HRSD24 middle insomnia item were associated with significantly more minutes awake in the middle of the night but the difference between the mean number of minutes awake in the middle of the night from the sleep diary that correspond to ratings of 0 and 1 was small (5 min) and not clinically meaningful. Sleep efficiencies, derived from the sleep diary (see Methods), were significantly and negatively correlated with the insomnia scale score of the HRSD24 (q = 0.51, df = 395, p < 0.0001). When using the conventional clinically significant cutoff of 85% sleep efficiency to split the sample into high and low groups SE groups (Morin et al., 1999), there was a statistically significant difference in the HRSD24 insomnia scale score between the high and low SE groups (Mann–Whitney: z = 8.5, p < 0.0001). Fig. 1(a) depicts this difference.
485
2.0
1.5
1.0
0.5
0.0
-0.5 Low
High
Sleep Efficiency (b)
3
2
1
0
Low
HIgh
Fig. 1. (a) Mean of sleep items on HRSD24 for patients with high and low* sleep efficiency *high (n = 119) and low (n = 276) sleep efficiency (SE) groups were defined using the clinical cutoff of 85% (Morin et al., 1999). The groups are statistically different on the HRSD24 sleep items (Mann–Whitney: z = 8.5, p < .0001). (b) Mean of sleep items on IDSSR30 for patients with high and low* sleep efficiency high (n = 116) and low (n = 266) sleep efficiency (SE) groups were defined using the clinical cutoff of 85% (Mowbray, 1972). The groups are statistically different on the IDS-SR30 sleep items (Mann–Whitney: z = 8.4, p < .0001). Circles indicate outliers defined as 1.5–3 interquartile ranges above the median.
tests followed by Mann–Whitney comparisons confirm that the means in number of minutes to fall asleep, the number of minutes awake after sleep onset, and the number of minutes awake before planned were statistically significantly different between the four levels of
486
R. Manber et al. / Journal of Psychiatric Research 39 (2005) 481–488
each of the three insomnia items on the IDS-SR30 (p < 0.001). For early insomnia, increased scores on the IDS-SR30 sleep item were associated with significantly more minutes to fall asleep (p < 0.001). For late insomnia, the number of minutes awake before planned corresponding to the lowest and highest scores of late insomnia on the IDS-SR30 were significantly different than the corresponding ranking in the expected direction (p < 0.001), but the number of minutes awake before planned was not statistically significantly greater between the two intermediate ratings (1 and 2, p = 0.30). For middle insomnia, the mean number of minutes awake in the middle of the night was significantly different between IDS-SR30 ratings of 0 and 1 and between IDS-SR30 ratings of 2 and 3 (p < 0.001). However, contrary to expectations, the number of minutes awake in the middle of the night was statistically significantly greater for the IDS-SR30 rating of 1 compared with the rating of 2 (p = 0.024). Sleep efficiencies were significantly and negatively correlated with the insomnia scale score of the IDSSR30 (q = 0.56, df = 380, p < 0.0001). Again, using the conventional clinically significant cutoff of 85% sleep efficiency to split the sample into high and low SE groups, we observed a statistically significant difference on the IDS-SR30 insomnia scale score (Mann–Whitney: z = 8.4, p < 0.0001). Fig. 1(b) depicts this difference.
3. Discussion The present study of insomnia in chronic depression provides a methodological bridge between the fields of depression and sleep research for assessment of insomnia. The study demonstrates convergent validity between retrospective measures of disturbed sleep that are used in depression research and prospective measures of sleep disturbances that are derived from a week of daily sleep diary entries and are commonly used in sleep research. The results indicate moderate to highly significant correlations between the two sources of data and provide a translation between global ratings of difficulty sleeping provided by the HRSD24 and the IDS-SR30 sleep items and the corresponding number of minutes awake at the beginning, middle, and end of the night. The data also suggest that the global insomnia subscale scores of the HRSD24 and the IDS-SR30 are clinically meaningful in that they are consistent with the acceptable threshold of 85% sleep efficiency used in clinical insomnia trial to distinguish between individuals with and without insomnia (e.g., Lichstein et al., 2003). To be meaningfully interpretable, each unit increment in the score for a given sleep item on the depression instrument should be associated with an increase in number of unwanted minutes of wakefulness for the corresponding portion of the night as recorded in daily
logs. This was indeed the case for the three HRSD24 sleep items (score range from 0 to 2), but not uniformly so for all IDS-R items (score range from 0 to 3). Unit increments in the score of each HRSD24 insomnia item (early, middle, and late) were associated with a significant increase in the weekly average number of minutes to fall asleep, minutes spent awake in the middle of the night, and minutes awake before intending to waken that were recorded each morning by the participants. This was also the case for the four levels of early insomnia item of the IDS-SR30, but not consistently for its late and middle insomnia items. For these two items, an increase in score from 1 to 2 was not associated with significantly greater number of minutes of unwanted wakefulness. In fact, the number of minutes awake in the middle of the night was significantly lower for a score of 2 than a score of 1 on the middle insomnia item of the IDS-SR30. This poor discrimination most likely reflects the written anchors of the middle insomnia item, which do not clearly describe an increment in severity (1 = I have a restless, light sleep with a few brief awakenings each night; 2 = I wake up at least once a night, but I go back to sleep easily). Poor item discrimination was also evident for the two intermediate ratings of late insomnia. The number of minutes awake before planned was lower for a score of 2 compared to a score of 1 on the late insomnia item of the IDS-SR30, albeit not significantly so. This too is likely related to verbal anchors that do not clearly represent increasing severity (1 = More than half the time, I awaken more than 30 min before I need to get up; 2 = I almost always awaken at least one hour or so before I need to, but I go back to sleep eventually). The data suggest that a non-zero score of early insomnia on either depression assessment corresponds to a weekly average of at least 32 min to fall asleep. This threshold is similar to the 31 min cutoff severity criterion suggested recently by Lichstein et al. (2003) for defining sleep onset insomnia in clinical trials. These authors also suggest 31 min as a severity cutoff criterion for sleep maintenance insomnia, usually defined as the number of minutes of wakefulness between sleep onset and the time the individual leaves the bed. This definition makes no distinction between middle insomnia (number of minutes awake from sleep onset to the last awakening) and late insomnia (the number of minutes awake before intended, which is typically associated with lying in bed awake before actually leaving bed). This distinction precludes direct comparisons between the suggested threshold for sleep maintenance insomnia and the results of this study. In the context of depression, the distinction between middle and late insomnia matters. To the best of our knowledge, this is the first study in which prospective sleep diaries were used to estimate the severity of early morning insomnia in depression, expressed as the number of minutes awake before desired.
R. Manber et al. / Journal of Psychiatric Research 39 (2005) 481–488
Several limitations in our study suggest caution in generalizing from the findings. First, our sample was restricted to patients with chronic depression with a relatively severe level of depressive symptoms (HRSD24 score >20 with a mean sample score of 27). The extent to which the findings would generalize to those with milder or episodic depression is unclear. Second, sleep diary data were available for only 58% of the full sample. Participants who provided sleep diary data, and were therefore included in the sample of this study, had significantly more severe middle and late insomnia (as measured by the HRSD sleep items) than participants in the parent study who did not complete diary. The possibility that this source of bias impacted the level of agreement between sleep diaries and interview based sleep ratings cannot be ruled out. Third, although patients with many comorbid conditions including most Axis II disorders were included, a number of specific comorbid psychiatric disorders and unstable medical disorders were excluded, thus potentially constraining generalizability of these findings. Fourth, it is possible that daily recording of information about sleep in the sleep diary at baseline prior to the HRSD and IDS-SR baseline assessments might have artificially increased the concordance of their answers. Finally, we cannot rule out the possibility that some participants were not fully compliant and filled out diary data for all days just prior to turning it in (and temporally close to retrospective assessment of sleep on self-report measures). This would artificially increase agreement between the two measures It is important to emphasize that although the data suggest high agreement between HRSD derived and sleep-diary derived measures of sleep difficulties; it will be unwise to conclude that sleep measures derived from depression measures should become the standard means to assess sleep in depression. Sleep diaries provide quantitative data that is not possible to derive from sleep items in most depression inventories (e.g., perceived number of minutes spent in un-wanted wakefulness during the time spent in bed). In fact, as insomnia is being increasingly recognized as relevant to patientsÕ prognosis, it is important that treatment outcome studies collect more data on sleep. The field of sleep research has developed multiple subjective and objective measures of sleep continuity. The decision of which measure(s) to use will dependent largely on the main aim of the study. Consideration needs to be given to issues of burden to the participant and feasibility. The accuracy of prospective sleep logs can be enhanced by the use of technology that provides time stamps and automated checks for internal consistency of responses (e.g., personal digital assistant (PDA) and web based logs). Some retrospective sleep questionnaires also provide estimates of the distribution of sleep and wakefulness and have high agreement with prospective sleep
487
logs (e.g., Monk et al., 2003). Objective sleep measures are typically more expensive and are therefore more adequate in small studies with a specific focus on sleep as the main predictor or outcome measure. A complete review of the relative merits of each measure is beyond the scope of this manuscript. Please see (Spielman et al., 2003) for a review of issues related to the assessment of sleep.
Acknowledgement This work was supported by a grant from the BristolMyers Squibb Company.
References American Psychiatric Association. Diagnostic and statistical manual of mental disorders. American Psychiatric Association, 1994. Dew MA, Reynolds CF, Houck PR, et al. Temporal profiles of the course of depression during treatment. Predictors of pathways toward recovery in the elderly. Archieves of General Psychiatry 1997;54(11):1016–24. First MB, Spitzer RL, Gibbon M, et al. Structured clinical interview for DSM-IV axis I disorders – Patient edition SCID-I/P, Version 2.0. Biometrics Research Department, New York State Psychiatric Institute; 1995. First MB, Gibbon M, Spitzer RL, et al. Structured clinical interview for DSM-IV axis II disorders (SCID-II). American Psychiatric Press; 1997. Fleck MPA, Poirier-Littre MF, Guelfi J-D, et al. Factorial structure of the17-item Hamilton Depression Rating Scale. Acta Psychiatrica Scandinavica 1995;92:168–72. Hamilton M. A rating scale for depression. Journal of Neurology Neurosurgery and Psychiatry 1960;25:55–62. Keller MB, McCullough JP, Klein DN, et al. A comparison of nefazodone, cognitive behavioral analysis system of psychotherapy, and their combination for the treatment of chronic depression. NEJM 2000;342:1462–70. Lichstein KL, Durrence HH, Taylor DJ, et al. Quantitative criteria for insomnia. Behaviour Research and Therapy 2003;41(4): 427–45. Maes M, Meltzer HY, Scharpe S, et al. Psychomotor retardation, anorexia, weight loss, sleep disturbances, and loss of energy: psychopathological correlates of hyperhaptoglobinemia during major depression. Psychiatry Research 1993;47(3):229–41. Manber R, Rush AJ, Thase ME, et al. The effects of psychotherapy, nefazodone, and their combination on subjective sleep in chronic depression. Sleep 2003;15(2):130–6. McCall WV, Reboussin BA, Cohen W. Subjective measurement of insomnia and quality of life in depressed inpatients. Journal of Sleep Research 2000;9(1):43–8. McCullough JPJ. Treatment for chronic depression. Cognitive behavioral analysis system of psychotherapy. The Guilford Press; 2000. Monk TH, Buysse DJ, Kennedy KS, et al. Measuring sleep habits without using a diary: the sleep timing questionnaire. Sleep 2003;26(2):208–12. Morin CM, Colecchi C, Stone J, et al. Behavioral and pharmacological therapies for late-life insomnia: a randomized controlled trial. JAMA 1999;281:991–9. Mowbray RM. The Hamilton Scale for Depression: a factor analysis. Psychological Medicine 1972;2:272–80.
488
R. Manber et al. / Journal of Psychiatric Research 39 (2005) 481–488
Nierenberg AA, Keefe BR, Leslie VC, et al. Residual symptoms in depressed patients who respond acutely to fluoxetine. Journal of Clinical Psychiatry 1999;60(4):221–5. Ramos-Brieva JA, Cordero-Villafila A. The new validation of the Hamilton Rating Scale for Depression. Journal of Psychiatry Research 1988;22(1):21–8. Reynolds CF, Frank E, Houck PR, et al. Which elderly patients with remitted depression remain well with continued interpersonal psychotherapy after discontinuation of antidepressant medication?. American Journal of Psychiatry 1997;154(7):958–62. Rush AJ, Giles DE, Schlesser MA, et al. The Inventory for Depressive Symptomatology (IDS): preliminary findings. Psychiatry Research 1986;18:65–87. Rush AJ, Gullion CM, Basco MR, et al. The Inventory of Depressive Symptomatology IDS: psychometric properties. Psychological Medicine 1996;26:477–86. Spielman AJ, Conroy D, Glovinsky PB. Evaluation of insomnia. In: Perlis ML, Lichstein KL, editors. Treating sleep disorders: principles and practice of behavioral sleep medicine. Berlin: Wiley; 2003.
Thase ME, Jindal RD. Combining psychotherapy and psychopharmacology for treatment of mental disorders. In: Bergin AE, Garfield SL, editors. Handbook of psychotherapy and behavior change. Wiley; 2002. Thase ME, Simons AD, Reynolds 3rd CF. Abnormal electroencephalographic sleep profiles in major depression: association with response to cognitive behavior therapy. Archieves of General Psychiatry 1996;53(2):99–108. Thase ME, Buysse DJ, Frank E, et al. Which depressed patients will respond to interpersonal psychotherapy. The role of abnormal EEG sleep profiles. American Journal of Psychiatry 1997;154(4):502–9. Thase ME, Rush AJ, Manber R, et al. Effects of nefazodone and cognitive behavioral analysis system of pasycotherapy, singly and in combination, on insomnia associated with chronic depression. Journal of Clinical Psychiatry 2002;63(6):493–500. Van Londen L, Molenaar RP, Goekoop JG, et al. Three- to 5-year prospective follow-up of outcome in major depression. Psychological Medicine 1998;28(3):731–5.
Journal of Psychiatric Research 39 (2005) 481–488
www.elsevier.com/locate/jpsychires
Assessing insomnia severity in depression: comparison of depression rating scales and sleep diaries Rachel Manber a,*, Christine Blasey a, Bruce Arnow a, John C. Markowitz b, Michael E. Thase c, A. John Rush d, Frank Dowling e, James Koscis b, Madhukar Trivedi d, Martin B. Keller f a
c
Department of Psychiatry and Behavioral Sciences, Stanford University, 401 Quarry Road, Suite 3301, Stanford, CA 94305 650, USA b Department of Psychiatry, Cornel University Medical College, New York, NY, USA Department of Psychiatry, University of Pittsburgh School of medicine, Western Psychiatric Institute and Clinics, Pittsburgh, PA, USA d Department of Psychiatry, University of Texas Southwestern Medical Center, Dallas, TX, USA e Department of Psychology, State University of New York at Stony Brook, Stony Brook, NY, USA f Department of Psychiatry and Behavioral Sciences, Butler Hospital, Brown University, Providence, RI, USA Received 21 July 2004; received in revised form 17 November 2004; accepted 3 December 2004
Abstract Depression and sleep researchers typically assess insomnia severity differently. Whereas depression researchers usually assess insomnia with items on depression symptom inventories, sleep researchers usually assess the subjective experience of insomnia with sleep diaries. The present manuscript utilizes baseline data from 397 participants in a large multi-site chronic depression study to assess agreement between these two methodologies. The results indicate that the early, middle, and late insomnia items of the Hamilton Rating Scale for Depression (HRSD24) and the Inventory of Depression Symptoms – Self Report (IDS-SR30) are highly correlated with the weekly mean values of time to sleep onset, time awake after sleep onset, and time awake prior to the planned wake-up obtained from prospective sleep diaries. Results also reveal significant correspondence between the weekly-mean of daily sleep efficiency, an accepted measure of sleep continuity (the ratio between reported time asleep and time in bed), and the insomnia scale scores of the HRSD24 and the IDS-SR30 (the mean score on the three insomnia items of each depression measure). Unit increments in HRSD24 scores for early, middle and late insomnia were associated with significant increases in unwanted minutes awake for corresponding periods on sleep diaries. Similar relationships were found for early insomnia on the IDS-SR30 but not for middle and late insomnia. Overall, with few exceptions, findings revealed substantial agreement between the HRSD24, IDS-SR30 and prospective sleep diary data. The study supports the validity of the sleep items and sleep subscales of the HRSD24 and the IDS-SR30 as global measures of insomnia severity in depression. Conventional sleep assessment procedures can complement depression scales by providing additional information about specific aspects of sleep in depression. Ó 2005 Elsevier Ltd. All rights reserved. Keywords: Early insomnia; Middle insomnia; Late insomnia; Depression; Assessment
Sleep disturbance is a common and key depressive symptom that affects both the course of illness and treat*
Corresponding author. Tel.: +1 650 724 2377; fax: +1 650 725 8910. E-mail address: [email protected] (R. Manber). 0022-3956/$ - see front matter Ó 2005 Elsevier Ltd. All rights reserved. doi:10.1016/j.jpsychires.2004.12.003
ment response. Patients with abnormal sleep profiles have demonstrated significantly worse outcome to psychotherapy than depressed patients with normal sleep profiles in terms of symptom severity, response, remission, and attrition (Thase et al., 1996, 1997). In addition, subjective complaints of insomnia appear to affect the
482
R. Manber et al. / Journal of Psychiatric Research 39 (2005) 481–488
stability of response to depression treatment with a combination of psychotherapy and pharmacotherapy and the time-course to relapse (Dew et al., 1997). Disturbed sleep is also the most common residual symptom following the acute phase of fluoxetine therapy of major depression, present in approximately 44% of treatment completers (Nierenberg et al., 1999). The presence of insomnia at the end of treatment predicts earlier and more frequent relapse (Reynolds et al., 1997; Van Londen et al., 1998). Despite the significance of insomnia, depression researchers typically do not assess insomnia as comprehensively as sleep researchers. In most insomnia studies, sleep disturbance is measured with a prospective selfreport method, such as a sleep diary, and/or with an objective measure, such as an all night polysomnographic study or a sensitive activity monitoring device from which data on sleep and wakefulness can be extracted. On the other hand, depression researchers abstract an index of insomnia severity from items on standardized questionnaires or structured clinical interviews used to assess depression severity. For example, some investigators use the average of the insomnia items of the Hamilton Rating Scale for Depression, (Hamilton, 1960) or the Inventory of Depression Symptoms – Self Report (IDS-SR, Rush et al., 1986, 1996) to track change in insomnia severity during treatment for depression (Maes et al., 1993; McCall et al., 2000; Thase et al., 2002). While the HRSD24 and IDS-SR30 have good internal consistency and test–retest reliability (RamosBrieva and Cordero-Villafila, 1988; Rush et al., 1996), the validity of their individual insomnia items and insomnia subscale have not been investigated. The present study is the first to evaluate agreement between retrospective measures of insomnia derived from items on the aforementioned depression inventories and a week of prospective sleep diary data. The sleep diary was selected as the primary sleep measure because it is inexpensive, relatively easy to administer in a large, multi-site trial, and it provides prospective data on participantsÕ perceived sleep. IndividualsÕ perceptions of their sleep provide clinically meaningful data, as it is the subjective perception of sleep that guides help-seeking behaviors. We previously reported that both the insomnia scales of depression inventories (the HRSD24 the IDS-SR30) and measures derived from a sleep diary were sensitive to differential changes in perceived sleep difficulties associated with three treatment modalities, medication, psychotherapy, and their combination (Manber et al., 2003; Thase and Jindal, 2002). In the present manuscript we present findings on the relationship between the weekly mean values of time to sleep onset, time awake after sleep onset, and time awake prior to planned rise time, and the corresponding insomnia items of the IDS-SR30 and the HRSD24 (early, middle, and late insomnia). This
report also examines the relationship between sleep efficiency, a global measure of disturbed sleep that is commonly derived from the sleep diary (defined below), and global measures of insomnia severity derived from the two depression measures.
1. Method 1.1. Design Sleep data were collected as part of a large 12-site, randomized, chronic depression study that compared the efficacy of Cognitive Behavioral Analysis System of Psychotherapy (CBASP), a structured, short-term psychotherapy developed specifically for chronic depression (McCullough, 2000), the antidepressant medication nefazodone, and the combination of CBASP and nefazodone for patients with chronic depression (Keller et al., 2000). For a comprehensive description of the methods employed in study, we refer the reader to Keller et al. (2000). This report is based on data collected at baseline from participants who provided sleep data on the three HRSD24 and IDS-SR30 sleep items (i.e., early, middle, and late insomnia) and on the three corresponding sleep diary items (i.e., time to fall asleep time awake, in the middle of the night, and time awake before planned). 1.2. Participants The current study sample, the Sleep Diary Sample consisted of 397 participants who completed sleep diary data prior to randomization. The sample constitutes 58% of the randomized sample of the parent study (n = 681) and had virtually identical demographic characteristics, shown in Table 1 (Keller et al., 2000) 1. Participants were adult outpatients (66.8% female) between the ages of 18 and 75 (mean age 44.2 years, SD = 10.9 years). Eligible participants met Diagnostic and Statistical Manual of Mental Disorders, fourth edition (American Psychiatric Association, 1994) criteria for a current chronic major depressive episode that was either at least 2 years of duration or was superimposed on a preexisting dysthymic disorder, or was the most recent major depressive episode (at least 2 years duration) in a prior recurrent MDD with incomplete inter-episode recovery. Eligibility for the larger study, and hence for the current sample, required an age between 18 and 75 1 There were, however, statistically significant differences in HRSD middle and late insomnia items with those who completed the sleep diary (the present sample) having higher scores than those who did not complete the sleep diary and were therefore not included in the study (1.2 ± 0.8 versus 1.4 ± 0.8 for middle insomnia and 0.9 ± 0.9 versus 1.0 ± 0.9 for late insomnia p < .05).
R. Manber et al. / Journal of Psychiatric Research 39 (2005) 481–488
483
Table 1 Demographic characteristics of current (n = 397) and original sample (n = 681)
Age (years) Female gender (%) White race (%) Marital status (%) Married/cohabitating Single Widowed Divorced/separated Depression diagnosis (%) Chronic major depression MDD with dysthymic disorder Recurrent depression, with incomplete Inter-episode recovery HRSD24 score ± SD HRSD early insomnia score ± SD HRSD middle insomnia score ± SD HRSD late insomnia score ± SD
Sleep diary sample N = 397
RCT sample N = 681
44.2 ± 10.9 66.8 91.7
43.0 ± 10.7 65.3 90.5
47.3 24.4 2.3 25.9
42.7 27.2 2.1 28.0
33.2 45.1 21.7
35.1 42.3 22.6
26.9 ± 5.0 0.92 ± 0.89 1.41 ± 0.80 1.02 ± 0.89
26.9 ± 5.0 0.92 ± 0.89 1.3 ± 0.82 0.95 ± 0.89
and a score of at least 20 on the 24-item HRSD. Participants were excluded for the following reasons: a history of psychotic, bipolar or obsessive compulsive disorder; eating disorders within the past year; substance abuse or dependence within the past six months; a high risk for suicide; antisocial, schizotypal, or severe borderline (i.e., high risk for hospitalization or significant self-mutilation) personality disorder; dementia; seizure disorder; poorly controlled or serious medical disorders; and a history of failing adequate trials of two different types of antidepressant medications or two different courses of empirically supported psychotherapy for depression within the past 3 years. Diagnoses were derived using a modified version of the Structured Clinical Interview for the DSM-IV-R (First et al., 1995) for Axis I disorders and an abbreviated version of the SCID-II (First et al., 1997) for Axis II disorders (limited to antisocial, borderline, schizotypal, avoidant, dependent, and obsessive–compulsive personality disorders). All participants had a physical examination, routine laboratory tests, and urine toxicology screen. All patients provided written informed consent, approved by the relevant institutionÕs human subjects committee before entering the study. 1.3. Measures 1.3.1. Sleep diary data Sleep diaries were completed daily (prospectively) for one week before treatment began and before the baseline assessments of depression severity. Diary entries included bedtime, lights out time, actual wake-up time, planned wake-up time, last time out of bed, latency to sleep onset (minutes), and minutes of wakefulness after sleep onset (WASO). The following variables were computed from the diary entries: (1) time in bed (TIB), defined as the time
elapsed from bedtime until the last time out of bed; (2) total sleep time (TST), defined as the time elapsed from bed time until wake up time minus latency to sleep onset and WASO; (3) sleep efficiency (SE), defined as the ratio of TST to TIB; and (4) early morning awakening (EMA), defined as the difference between actual wake up time and planned wake up time; a negative value for EMA reflects an actual wake up time that is earlier than the planned wake up time. 1.3.2. HRSD Insomnia items The 24-item HRSD, an observer-rated measure of depression severity, includes three items that assess disturbed sleep. Although factor analytic studies of the HRSD24 have yielded inconsistent results, the three insomnia items of the HRSD24 (early, middle, and late insomnia), each rated by the interviewer on a scale ranging from 0 (absent) to 2 (severe), frequently load on the same HRSD24 factor (Fleck et al., 1995; Mowbray, 1972; Ramos-Brieva and Cordero-Villafila, 1988). The insomnia scale score is computed as the average of the three insomnia items, and therefore has a range between 0 and 2. The HRSD insomnia scale score can detect improvement in insomnia associated with effective treatment of depression (Thase et al., 2002) and has been shown to be sensitive to the differential impact of psychotherapy and antidepressant medications on sleep (Thase et al., 2002). In the present study, HRSD raters received external certification of competence and were unaware of patientsÕ treatment conditions. 1.3.3. IDS-SR insomnia items The 30-item IDS-SR is a self-rating scale of depression severity and includes three insomnia items (early, middle, and late insomnia), each rated on a 0–3 scale. The early insomnia item is anchored as follows: 0 = I never take longer than 30 min to fall asleep; 1 = I take
484
R. Manber et al. / Journal of Psychiatric Research 39 (2005) 481–488
Table 2 Spearman correlations between HRSD24 Insomnia items and sleep diary data (n = 397) and between IDS-SR30 insomnia items and sleep diary data (n = 384) Sleep diary
Early insomnia
Middle insomnia
Late insomnia
HRSD24 insomnia items Minutes to fall asleep Minutes awake after sleep onset Minutes awake before planned
+0.63 +0.16 0.01
+0.17 +0.49 0.26
+0.11 +0.26 0.49
IDS-SR30 insomnia items Minutes to fall asleep Minutes awake after sleep onset Minutes awake before planned
+0.70 +0.26 +0.02
+0.21 +0.60 0.29
+0.13 +0.31 0.46
at least 30 min to fall asleep, less than half the time; 2 = I take at least 30 min to fall asleep, more than half the time; and 3 = I take more than 60 min to fall asleep, more than half the time. The middle insomnia item is anchored as follows: 0 = I do not wake up at night; 1 = I have a restless, light sleep with a few brief awakenings each night; 2 = I wake up at least once a night, but I go back to sleep easily; 3 = I awaken more than once a night and stay awake for 20 min or more, more than half the time. The late insomnia item is anchored as follows: 0 = Most of the time, I awaken no more than 30 min before I need to get up; 1 = More than half the time, I awaken more than 30 min before I need to get up; 2 = I almost always awaken at least 1 h or so before I need to, but I go back to sleep eventually; 3 = I awaken at least 1 h before I need to, and cannot go back to sleep. The insomnia scale score is computed as the average of the three insomnia items, and therefore has a range be-
tween 0 and 3. The IDS-SR30 insomnia scale score can detect improvement in insomnia associated with effective treatment of depression and it has demonstrated sensitivity to the differential impact of psychotherapy and antidepressant medications on sleep (Thase et al., 2002).
2. Results 2.1. Concordance of sleep diary data and HRSD24 Moderately statistically significant Spearman correlations, shown on the diagonal of Table 2, were observed between HRSD24 and corresponding sleep diary responses. Off diagonal entries in Table 2 are correlations between HRSD24 and non-corresponding sleep diary items. Comparisons of off-diagonal entries in Table 2
Table 3 Mean ± SD and (Median) time (minutes) from the sleep diary for each level of the corresponding item on the HRSD24 (n = 397) and the IDS-SR30 (n = 384)
HRSD24 0 1 2 Paired comparisons IDS-SR30 0 1 2 3 Paired comparisons a
Earlya
Middlea
Latea
Time to sleep onset
Time awake after sleep onset
Time awake before planned
21.6 ± 19.8 (16.2) 39.0 ± 23.4 (30.0) 67.2 ± 43.8 (58.8) 0<1<2
16.3 ± 24.1 (7.1) 21.2 ± 23.7 (13.5) 47.0 ± 39.9 (36.9) 0 < 1*, 1 < 2
17.4 ± 78.0 ( 10.8) 17.4 ± 78.0 (16.2) 66.0 ± 90.0 (54.6) 0>1>2
17.8 + 19.7 (12.5) 31.6 + 21.6 (25.0) 49.4 + 29.3 (38.9) 84.7 + 45.8 (76.4) 0<1<2<3
3.5 + 4.5 (1.0) 26.4 + 25.6 (16.4) 17.7 + 16.2 (13.3) 54.8 + 32.2 (51.4) 0 < 1, 1 > 2*, 2 < 3
11.9 + 74.0 (6.4) 40.3 + 59.3 ( 32.5) 33.3 + 110.4 ( 28.2) 85.7 + 86.3 ( 73.6) 0 > 1, 1 = 2, 2 > 3
Kruskal–Wallis H test indicated significant differences in times by HRSD24 and IDS-SR30 rating for each insomnia item, all p values < 0.0001. Significant follow-up comparisons, using Mann–Whitney, are shown in the last row of each table, all p values, except the one marked with * < 0.001 (*p < 0.05). For late insomnia, a negative number represents a wakeup time that is earlier than planned.
R. Manber et al. / Journal of Psychiatric Research 39 (2005) 481–488
2.2. Concordance of sleep diary data and IDS-SR30 IDS-SR30 data were available for the majority of participants in the sample (n = 384, 97%). Shown on the diagonal of Table 2, Spearman correlations between sleep diary data and their corresponding IDS-SR30 items were +0.70 (early), +0.60 (middle), and 0.46 (late). Correlations off the diagonal were lower; effect sizes in terms of r2 were half of those observed on the diagonal, indicating substantial item discrimination. The mean and median times in minutes from sleep diary items for each level of the corresponding IDSSR30 item are depicted in Table 3. Kruskal–Wallis H
(a) 2.5
Mean of 3 HAMD Sleep Items
with those on the diagonal indicate substantial item discrimination, particularly for early insomnia. The HRSD24 early insomnia item was significantly and positively correlated with its corresponding item in the sleep diary, minutes to fall asleep (q = 0.63, df = 395, p < 0.0001). The HRSD24 middle insomnia item was significantly and positively correlated with diary recordings of minutes awake after sleep onset (q = 0.49, df = 395, p < 0.0001). The HRSD24 late insomnia item was inversely related to diary recordings of minutes awake before planned (q = 0.49, df = 365, p < 0.0001). The mean and median times in minutes from sleep diary items for each level of the corresponding HRSD24 item are depicted in Table 3. Kruskal–Wallis H tests followed by Mann–Whitney comparisons confirm that the means in number of minutes to fall asleep, minutes awake after sleep onset, and minutes awake before planned were statistically significantly different among the three levels of each of the three corresponding insomnia items on the HRSD24 (p < 0.001). For early and late insomnia increased scores on the HRSD24 were associated with significantly more minutes to fall asleep and with significantly more minutes awake in the morning before planned, respectively (all pÕs < 0.001). Increased scores on the HRSD24 middle insomnia item were associated with significantly more minutes awake in the middle of the night but the difference between the mean number of minutes awake in the middle of the night from the sleep diary that correspond to ratings of 0 and 1 was small (5 min) and not clinically meaningful. Sleep efficiencies, derived from the sleep diary (see Methods), were significantly and negatively correlated with the insomnia scale score of the HRSD24 (q = 0.51, df = 395, p < 0.0001). When using the conventional clinically significant cutoff of 85% sleep efficiency to split the sample into high and low groups SE groups (Morin et al., 1999), there was a statistically significant difference in the HRSD24 insomnia scale score between the high and low SE groups (Mann–Whitney: z = 8.5, p < 0.0001). Fig. 1(a) depicts this difference.
485
2.0
1.5
1.0
0.5
0.0
-0.5 Low
High
Sleep Efficiency (b)
3
2
1
0
Low
HIgh
Fig. 1. (a) Mean of sleep items on HRSD24 for patients with high and low* sleep efficiency *high (n = 119) and low (n = 276) sleep efficiency (SE) groups were defined using the clinical cutoff of 85% (Morin et al., 1999). The groups are statistically different on the HRSD24 sleep items (Mann–Whitney: z = 8.5, p < .0001). (b) Mean of sleep items on IDSSR30 for patients with high and low* sleep efficiency high (n = 116) and low (n = 266) sleep efficiency (SE) groups were defined using the clinical cutoff of 85% (Mowbray, 1972). The groups are statistically different on the IDS-SR30 sleep items (Mann–Whitney: z = 8.4, p < .0001). Circles indicate outliers defined as 1.5–3 interquartile ranges above the median.
tests followed by Mann–Whitney comparisons confirm that the means in number of minutes to fall asleep, the number of minutes awake after sleep onset, and the number of minutes awake before planned were statistically significantly different between the four levels of
486
R. Manber et al. / Journal of Psychiatric Research 39 (2005) 481–488
each of the three insomnia items on the IDS-SR30 (p < 0.001). For early insomnia, increased scores on the IDS-SR30 sleep item were associated with significantly more minutes to fall asleep (p < 0.001). For late insomnia, the number of minutes awake before planned corresponding to the lowest and highest scores of late insomnia on the IDS-SR30 were significantly different than the corresponding ranking in the expected direction (p < 0.001), but the number of minutes awake before planned was not statistically significantly greater between the two intermediate ratings (1 and 2, p = 0.30). For middle insomnia, the mean number of minutes awake in the middle of the night was significantly different between IDS-SR30 ratings of 0 and 1 and between IDS-SR30 ratings of 2 and 3 (p < 0.001). However, contrary to expectations, the number of minutes awake in the middle of the night was statistically significantly greater for the IDS-SR30 rating of 1 compared with the rating of 2 (p = 0.024). Sleep efficiencies were significantly and negatively correlated with the insomnia scale score of the IDSSR30 (q = 0.56, df = 380, p < 0.0001). Again, using the conventional clinically significant cutoff of 85% sleep efficiency to split the sample into high and low SE groups, we observed a statistically significant difference on the IDS-SR30 insomnia scale score (Mann–Whitney: z = 8.4, p < 0.0001). Fig. 1(b) depicts this difference.
3. Discussion The present study of insomnia in chronic depression provides a methodological bridge between the fields of depression and sleep research for assessment of insomnia. The study demonstrates convergent validity between retrospective measures of disturbed sleep that are used in depression research and prospective measures of sleep disturbances that are derived from a week of daily sleep diary entries and are commonly used in sleep research. The results indicate moderate to highly significant correlations between the two sources of data and provide a translation between global ratings of difficulty sleeping provided by the HRSD24 and the IDS-SR30 sleep items and the corresponding number of minutes awake at the beginning, middle, and end of the night. The data also suggest that the global insomnia subscale scores of the HRSD24 and the IDS-SR30 are clinically meaningful in that they are consistent with the acceptable threshold of 85% sleep efficiency used in clinical insomnia trial to distinguish between individuals with and without insomnia (e.g., Lichstein et al., 2003). To be meaningfully interpretable, each unit increment in the score for a given sleep item on the depression instrument should be associated with an increase in number of unwanted minutes of wakefulness for the corresponding portion of the night as recorded in daily
logs. This was indeed the case for the three HRSD24 sleep items (score range from 0 to 2), but not uniformly so for all IDS-R items (score range from 0 to 3). Unit increments in the score of each HRSD24 insomnia item (early, middle, and late) were associated with a significant increase in the weekly average number of minutes to fall asleep, minutes spent awake in the middle of the night, and minutes awake before intending to waken that were recorded each morning by the participants. This was also the case for the four levels of early insomnia item of the IDS-SR30, but not consistently for its late and middle insomnia items. For these two items, an increase in score from 1 to 2 was not associated with significantly greater number of minutes of unwanted wakefulness. In fact, the number of minutes awake in the middle of the night was significantly lower for a score of 2 than a score of 1 on the middle insomnia item of the IDS-SR30. This poor discrimination most likely reflects the written anchors of the middle insomnia item, which do not clearly describe an increment in severity (1 = I have a restless, light sleep with a few brief awakenings each night; 2 = I wake up at least once a night, but I go back to sleep easily). Poor item discrimination was also evident for the two intermediate ratings of late insomnia. The number of minutes awake before planned was lower for a score of 2 compared to a score of 1 on the late insomnia item of the IDS-SR30, albeit not significantly so. This too is likely related to verbal anchors that do not clearly represent increasing severity (1 = More than half the time, I awaken more than 30 min before I need to get up; 2 = I almost always awaken at least one hour or so before I need to, but I go back to sleep eventually). The data suggest that a non-zero score of early insomnia on either depression assessment corresponds to a weekly average of at least 32 min to fall asleep. This threshold is similar to the 31 min cutoff severity criterion suggested recently by Lichstein et al. (2003) for defining sleep onset insomnia in clinical trials. These authors also suggest 31 min as a severity cutoff criterion for sleep maintenance insomnia, usually defined as the number of minutes of wakefulness between sleep onset and the time the individual leaves the bed. This definition makes no distinction between middle insomnia (number of minutes awake from sleep onset to the last awakening) and late insomnia (the number of minutes awake before intended, which is typically associated with lying in bed awake before actually leaving bed). This distinction precludes direct comparisons between the suggested threshold for sleep maintenance insomnia and the results of this study. In the context of depression, the distinction between middle and late insomnia matters. To the best of our knowledge, this is the first study in which prospective sleep diaries were used to estimate the severity of early morning insomnia in depression, expressed as the number of minutes awake before desired.
R. Manber et al. / Journal of Psychiatric Research 39 (2005) 481–488
Several limitations in our study suggest caution in generalizing from the findings. First, our sample was restricted to patients with chronic depression with a relatively severe level of depressive symptoms (HRSD24 score >20 with a mean sample score of 27). The extent to which the findings would generalize to those with milder or episodic depression is unclear. Second, sleep diary data were available for only 58% of the full sample. Participants who provided sleep diary data, and were therefore included in the sample of this study, had significantly more severe middle and late insomnia (as measured by the HRSD sleep items) than participants in the parent study who did not complete diary. The possibility that this source of bias impacted the level of agreement between sleep diaries and interview based sleep ratings cannot be ruled out. Third, although patients with many comorbid conditions including most Axis II disorders were included, a number of specific comorbid psychiatric disorders and unstable medical disorders were excluded, thus potentially constraining generalizability of these findings. Fourth, it is possible that daily recording of information about sleep in the sleep diary at baseline prior to the HRSD and IDS-SR baseline assessments might have artificially increased the concordance of their answers. Finally, we cannot rule out the possibility that some participants were not fully compliant and filled out diary data for all days just prior to turning it in (and temporally close to retrospective assessment of sleep on self-report measures). This would artificially increase agreement between the two measures It is important to emphasize that although the data suggest high agreement between HRSD derived and sleep-diary derived measures of sleep difficulties; it will be unwise to conclude that sleep measures derived from depression measures should become the standard means to assess sleep in depression. Sleep diaries provide quantitative data that is not possible to derive from sleep items in most depression inventories (e.g., perceived number of minutes spent in un-wanted wakefulness during the time spent in bed). In fact, as insomnia is being increasingly recognized as relevant to patientsÕ prognosis, it is important that treatment outcome studies collect more data on sleep. The field of sleep research has developed multiple subjective and objective measures of sleep continuity. The decision of which measure(s) to use will dependent largely on the main aim of the study. Consideration needs to be given to issues of burden to the participant and feasibility. The accuracy of prospective sleep logs can be enhanced by the use of technology that provides time stamps and automated checks for internal consistency of responses (e.g., personal digital assistant (PDA) and web based logs). Some retrospective sleep questionnaires also provide estimates of the distribution of sleep and wakefulness and have high agreement with prospective sleep
487
logs (e.g., Monk et al., 2003). Objective sleep measures are typically more expensive and are therefore more adequate in small studies with a specific focus on sleep as the main predictor or outcome measure. A complete review of the relative merits of each measure is beyond the scope of this manuscript. Please see (Spielman et al., 2003) for a review of issues related to the assessment of sleep.
Acknowledgement This work was supported by a grant from the BristolMyers Squibb Company.
References American Psychiatric Association. Diagnostic and statistical manual of mental disorders. American Psychiatric Association, 1994. Dew MA, Reynolds CF, Houck PR, et al. Temporal profiles of the course of depression during treatment. Predictors of pathways toward recovery in the elderly. Archieves of General Psychiatry 1997;54(11):1016–24. First MB, Spitzer RL, Gibbon M, et al. Structured clinical interview for DSM-IV axis I disorders – Patient edition SCID-I/P, Version 2.0. Biometrics Research Department, New York State Psychiatric Institute; 1995. First MB, Gibbon M, Spitzer RL, et al. Structured clinical interview for DSM-IV axis II disorders (SCID-II). American Psychiatric Press; 1997. Fleck MPA, Poirier-Littre MF, Guelfi J-D, et al. Factorial structure of the17-item Hamilton Depression Rating Scale. Acta Psychiatrica Scandinavica 1995;92:168–72. Hamilton M. A rating scale for depression. Journal of Neurology Neurosurgery and Psychiatry 1960;25:55–62. Keller MB, McCullough JP, Klein DN, et al. A comparison of nefazodone, cognitive behavioral analysis system of psychotherapy, and their combination for the treatment of chronic depression. NEJM 2000;342:1462–70. Lichstein KL, Durrence HH, Taylor DJ, et al. Quantitative criteria for insomnia. Behaviour Research and Therapy 2003;41(4): 427–45. Maes M, Meltzer HY, Scharpe S, et al. Psychomotor retardation, anorexia, weight loss, sleep disturbances, and loss of energy: psychopathological correlates of hyperhaptoglobinemia during major depression. Psychiatry Research 1993;47(3):229–41. Manber R, Rush AJ, Thase ME, et al. The effects of psychotherapy, nefazodone, and their combination on subjective sleep in chronic depression. Sleep 2003;15(2):130–6. McCall WV, Reboussin BA, Cohen W. Subjective measurement of insomnia and quality of life in depressed inpatients. Journal of Sleep Research 2000;9(1):43–8. McCullough JPJ. Treatment for chronic depression. Cognitive behavioral analysis system of psychotherapy. The Guilford Press; 2000. Monk TH, Buysse DJ, Kennedy KS, et al. Measuring sleep habits without using a diary: the sleep timing questionnaire. Sleep 2003;26(2):208–12. Morin CM, Colecchi C, Stone J, et al. Behavioral and pharmacological therapies for late-life insomnia: a randomized controlled trial. JAMA 1999;281:991–9. Mowbray RM. The Hamilton Scale for Depression: a factor analysis. Psychological Medicine 1972;2:272–80.
488
R. Manber et al. / Journal of Psychiatric Research 39 (2005) 481–488
Nierenberg AA, Keefe BR, Leslie VC, et al. Residual symptoms in depressed patients who respond acutely to fluoxetine. Journal of Clinical Psychiatry 1999;60(4):221–5. Ramos-Brieva JA, Cordero-Villafila A. The new validation of the Hamilton Rating Scale for Depression. Journal of Psychiatry Research 1988;22(1):21–8. Reynolds CF, Frank E, Houck PR, et al. Which elderly patients with remitted depression remain well with continued interpersonal psychotherapy after discontinuation of antidepressant medication?. American Journal of Psychiatry 1997;154(7):958–62. Rush AJ, Giles DE, Schlesser MA, et al. The Inventory for Depressive Symptomatology (IDS): preliminary findings. Psychiatry Research 1986;18:65–87. Rush AJ, Gullion CM, Basco MR, et al. The Inventory of Depressive Symptomatology IDS: psychometric properties. Psychological Medicine 1996;26:477–86. Spielman AJ, Conroy D, Glovinsky PB. Evaluation of insomnia. In: Perlis ML, Lichstein KL, editors. Treating sleep disorders: principles and practice of behavioral sleep medicine. Berlin: Wiley; 2003.
Thase ME, Jindal RD. Combining psychotherapy and psychopharmacology for treatment of mental disorders. In: Bergin AE, Garfield SL, editors. Handbook of psychotherapy and behavior change. Wiley; 2002. Thase ME, Simons AD, Reynolds 3rd CF. Abnormal electroencephalographic sleep profiles in major depression: association with response to cognitive behavior therapy. Archieves of General Psychiatry 1996;53(2):99–108. Thase ME, Buysse DJ, Frank E, et al. Which depressed patients will respond to interpersonal psychotherapy. The role of abnormal EEG sleep profiles. American Journal of Psychiatry 1997;154(4):502–9. Thase ME, Rush AJ, Manber R, et al. Effects of nefazodone and cognitive behavioral analysis system of pasycotherapy, singly and in combination, on insomnia associated with chronic depression. Journal of Clinical Psychiatry 2002;63(6):493–500. Van Londen L, Molenaar RP, Goekoop JG, et al. Three- to 5-year prospective follow-up of outcome in major depression. Psychological Medicine 1998;28(3):731–5.