- Original research
- Open Access
Retrospective cross-validation of automated sleep staging using electroocular recording in patients with and without sleep disordered breathing
© Levendowski et al.; licensee BioMed Central Ltd. 2012
- Received: 3 March 2012
- Accepted: 24 May 2012
- Published: 25 June 2012
Alterations of sleep duration and architecture have been associated with increased morbidity and mortality, and specifically linked to chronic cardiovascular disease and psychiatric disorders, such as type 2 diabetes or depression. Measurement of sleep quality to assist in the diagnosis or treatment of these diseases is not routinely performed due to the complexity and cost of conventional methods. The objective of this study is to cross-validate the accuracy of an automated algorithm that stages sleep from the EEG signal acquired with sensors that can be self-applied by patients.
This retrospective study design included polymsomnographic records from 19 presumably healthy individuals and 68 patients suspected of having sleep disordered breathing (SDB). Epoch-by-epoch comparisons were made between manual vs. automated sleeps staging (from the left and right electrooculogram) with the impact of SDB severity considered.
Both scoring methods reported decreased Stage N3 and REM and increased wake and N1 as SDB severity increased. Inter-class correlations and Kappa coefficients were strong across all stages except N1. Agreements across all epochs for subjects with normal and patients with mild SDB were: wake = 80%, N1 = 25%, N2 = 78%, N3 = 84% and REM = 75%. Agreement decreased in patients with moderate and severe SDB amounting to: wake = 71%, N1 = 30%, N2 = 71%, N3 = 65%, and REM = 67%. Differences in detection of sleep onset were within three-minutes in 48 % of the subjects and 10-min in 73 % of the cases and were not impacted by SDB severity. Automated staging slightly underestimated total sleep time but this difference had a limited impact on the respiratory disturbance indexes.
This cross-validation study demonstrated that measurement of sleep architecture obtained from a single-channel of forehead EEG can be equivalent to between-rater agreement using conventional manual scoring. The accuracies obtained with automated sleep staging were inversely proportional to SDB severity at a rate similar to manual scorers. These results suggest that the automated sleep staging used in this study may prove useful in evaluating sleep quality in patients with chronic diseases.
Adequate amounts and quality of sleep are essential for health and well-being. Both short and long sleep durations are significant predictors of morbidity and all-cause mortality . Short duration sleep (< 6 hours per night) represents an independent risk factor for development of type 2 diabetes [2–7], central obesity (in women)  and psychiatric disorders such as depression, attention deficit and substance abuse [9–14]. A lack of certain phases of sleep may have adverse effects on health in spite of the seemingly adequate sleep duration. For example, an insufficient amount of slow wave sleep has been associated with hypertension , type 2 diabetes [3, 16] and increased risk of obesity , while anomalies of rapid eye movement (REM) sleep have been linked to dementia, depression and post-traumatic stress disorder (PTSD) [18–20].
Conventionally, the assessment of sleep architecture has been done in dedicated facilities and relied on multichannel polysomnography (PSG) and manual scoring of the data. PSG provides comprehensive information about sleep duration and architecture but it is too expensive and cumbersome for large-scale or repeated-measures evaluations. Manual scoring is a laborious practice even with the use of software to assist with scorers , and furthermore it is subject to considerable disagreement between experts in regard to assigned sleep stages, cumulative measures of sleep structure and indices of respiratory or other disturbances [22–24]. On the other hand, inexpensive tools that are validated and applicable on a large scale (i.e., wrist actigraphy and sleep diaries) provide only rudimentary estimates of total sleep time without any information about the quality of periods self-reported or labeled by actigraphy as sleep. As a result, sleep is rarely evaluated in patients with suspect or confirmed chronic disorders where its assessment might provide important clues for diagnosis or treatment.
Recent advances in electronic technologies and sensor interfaces have allowed for a significant reduction of the size and weight of recording equipment and made its self-application feasible. This would allow assessment of sleep quality in the home where a patient’s sleep patterns can be objectively quantified in their normal sleeping environment using wearable recorders with only few EEG electrodes below the hairline. A compelling advantage of such an approach is that adhesive ECG-type electrodes or pads made from conductive fabrics can be easily applied to the forehead or around the eyes by patients following a simple set of instructions. In recent years several algorithms for automated staging of sleep from a limited number of channels have been introduced and successfully validated in samples composed mostly or exclusively of healthy volunteers [25–27]. However similar levels of scoring accuracy need to be demonstrated in relevant clinical populations in order to support routine adoption.
This paper presents validation of a novel algorithm that stages sleep in conventional 30-second epochs from a single EEG channel. The algorithm was developed and initially validated in healthy subjects using a differential recording from two forehead electrodes (Fp1-Fp2) . In this study the validation has been extended to a clinical population composed mostly of subjects evaluated with laboratory PSG for suspected sleep-disordered breathing. As the frontopolar electrodes are not routinely applied during such evaluations, the differential input for the algorithm was derived from the left and right electrooculographic (EOG) channels.
Eighty-seven nocturnal polysomnographic records selected for this study were acquired under IRB review at the New York University Sleep Disorders Center. Prior to further stratification by sleep disordered breathing severity, the data set included 19 subjects presumed to be healthy and the balance suspected of having sleep disordered breathing (SDB). The ethnic profile included 52% white, 10% Asian, 9% African American, 9% Hispanic/Latino, and 20% unreported.
Distribution of group data based on sleep disordered breathing severity
n (% male)
Age, mean yrs ± SD
39 ± 14.6
41 ± 11.4
53 ± 15.8
47 ± 12.1
AHI, mean ± SD
1 ± 1.3
4 ± 2.7
14 ± 8.7
57 ± 24.4
RDI, mean ± SD
6 ± 2.3
15 ± 3.1
31 ± 6.7
71 ± 19.3
Minor modifications have been made to the algorithm prior to its application to this study’s data. As the frontopolar differential EEG derivation was not available in this study, the algorithm was applied to the differential signal derived by subtracting the left EOG from the right EOG channel. As the filters in the sleep staging algorithms are designed for a sampling rate of 256 Hz, the difference EOG signal, originally available at 128 Hz, was subsequently re-sampled at 256 Hz using linear interpolation. Finally, the EMG power threshold was reduced by two thirds and the BEI threshold was scaled up by a factor of three in order to reflect the reduced width of the effective EMG band available in this study (32–64 Hz as compared to 32–128 Hz in the original data set ). The validity of these modifications was verified by an ad-hoc analysis of 19 records from the healthy subjects and 18 studies with manually staged total sleep time less than 3.5 hours. It is important, however, to stress that the vast majority of the algorithm’s thresholds (21 out of 23) were not changed.
Inter-class correlations were used to assess concordance across subjects between the total time spent in each of the stages as determined by manual vs. auto-scoring. Cross-tabulations were made between the manual and automated scoring for all epochs, and the overall and stage-specific percentage agreement and Kappa coefficients were calculated. Based on similarities in the scoring accuracy, the data from the subjects with normal and patients with mild SDB data were pooled into one and the data from patients with moderate and severe SDB into a second group. The variability of stage-specific agreement across subjects was analyzed on box-whisker plots for each of the five AASM sleep stages. Bland-Altman plots were used to assess differences between manual and automated estimates of total sleep time, sleep efficiency (ratio of the total sleep time and total recording time), and respiratory disturbance index (RDI) across all subjects. Sleep onset was identified from both manual and automated scoring (as three consecutive non-wake epochs) and the absolute differences between the two estimates were analyzed as a function of SDB severity and accuracy of detection of stages Wake and N1 in each particular subject.
Correlations across staged time
Inter-Class Correlation (ICC) between manual and auto-scored time by stage
Pooled accuracy by stage
Pair-wise epoch by epoch agreement by SDB group
Normal and Mild SDB (n = 40,641 epochs)
Moderate and Severe SDB (n = 29,938 epochs)
Based on manual scoring, the distribution of sleep architecture changed across the two groups as SDB severity increased, most noticeable with stage N2 (44.7 vs. 25.1%), stage N3 (10.4 vs. 2.7%) and REM (13.4 vs. 6.4%). There were also important differences in the agreement between manual and automated staging when the normal and mild SDB group was compared to the moderate and severe SDB group. The percentage of epochs identified as Wake by visual scoring and staged as N1 increased from 9.2% to 16.9% due to the influence of obstructive breathing on sleep continuity. Similarly, the percentage of epochs visually scored as N2 and staged as N1 increased across the two groups from 7.6% to 14.1%. The concordance for N3 decreased from 83.9% to 65.0%, with misclassifications shifting to N2 likely as a result of a decreased amount and continuity of slow wave sleep in patients with more severe SDB. Finally, the increase in REM sleep misclassified as N1 (10.8% vs. 6.6%) for those with moderate/severe SDB was likely influenced by the overall reduction in REM time.
Kappa coefficients showed strong agreement for Wake, N3 and REM and moderate agreement for N2 in the normal and mild SDB group. In patients with moderate and severe SDB, Kappa coefficients showed moderate agreement for Wake, N2, N3 and REM.
Staging accuracy by subject
TST and sleep efficiency
Distribution of subjects by SDB severity and sleep onset difference categories
≤ 3 min
4 to 10 min
11 to 20 min
> 20 min
This study evaluated the accuracy of a novel algorithm designed to stage sleep from a single EEG channel using two sensors on the forehead. Overall, the agreement between the manual and automated scoring for all stages decreased as the severity of SBD increased, but remained fairly high for stages Wake, N2 and REM. The agreement by stage across the normal/mild SDB group were remarkably similar. For the moderate/severe SDB group the overall agreement dropped as the algorithm tended to classify into ‘lighter’ stages as compared to the human scorer, i.e. scored N1 in place of stages N2, REM epochs as N1 or N2, and N2 in place of manually scored N3. The agreement between manual and automated scoring for stage N1 was generally poor irrespective of the SDB severity, and this stage was typically confused with Wake or N2. While the disagreement between stages N1 and N2 is not relevant for most clinical situations (where the two stages are often viewed as a single ‘light NREM’ stage), unbalanced misclassification of a large proportion of N1 epochs as Wake or vice versa could have an impact on total sleep time and RDI estimates.
The Bland-Altman plot of manual and automated estimates of total sleep time suggest the misclassifications of N1 as Wake were, on average, counterbalanced by misclassification of Wake as sleep, as the bias for TST estimates was negligible and variance clinically insignificant. The wake/sleep misclassifications were stronger as the SDB severity increased, in part because there were more wake epochs. The influence of the Wake/sleep misclassification on the automated estimates of RDI was, nonetheless, negligible.
The validity of an algorithm for automated sleep staging is often assessed by comparison of its overall and stage-specific accuracy to the agreement between two (or more) human raters who scored the same data. This study was limited by the fact that we compared accuracy to one set of sleep scores per patient, with several technicians contributing to the scoring of the data set. The approach reduced the bias that may have been introduced by a single rater, or the introduction of “invalid” epochs if two raters did not agree. Given our results were reliant on a single rater per study, the report by Norman and co-workers  provided a valuable benchmark for inter-rater agreement across normal subjects and patients with sleep disordered breathing. The agreement between our algorithm and manual scoring for normal/mild and moderate severe SDB groups (Table 2) were equivalent to the inter-rater reliability reported by Norman for Stages N2 and N3. Our auto-scoring vs. manual agreement was similar to inter-rater reliability for stage Wake in normal/mild SDB. Our agreement was slightly inferior to the reported inter-rater reliability for REM and N1, which may be partially explained by the fact that our SDB patients were more severe than the Norman cohort (RDI = 41 ± 28 vs. 34 ± 31 events/hr). Suboptimal recognition of cortical arousals resulting from SDB may also have contributed to the difficulties in the algorithm staging REM sleep and differentiation between wake and stage N1. The reduced EMG bandwidth and slightly different electrode placement might also be responsible, as the accuracy of both REM and N1 detection was better in the original validation study on healthy controls in which the EMG power was derived from 40 to 128 Hz.
For this study we used as the differential inputs signals from the left and right EOG, rather than a standard placement used for development and initial validation of the algorithm. The only adaptation made was to accommodate differences in sampling rates. Thus, this study served to assess the generalizability of the staging algorithm using forehead EEG in patients with sleep disordered breathing, a chronic disease known to disrupt sleep architecture and compromise manual staging accuracy. We made no effort to manually edit the full-disclosure auto-scored epochs, although the capability is provided for physicians boarded in sleep medicine to review and interpret the staging. One of the advantages of staging sleep from the forehead is access to information provided by eye movements. Manually editing would definitely have eliminated obvious misclassifications between stage N1 and REM resulting from slow eye rolls apparent after long periods of wake. The disadvantage is that alpha waves useful in staging N1 are visually undetectable.
Others algorithms have been evaluated for the staging of sleep from the forehead. The Kappa coefficients obtained for wake, light non-REM (N1 and N2), N3 (SWS) and REM in the test group of 131 subjects using two EOG channels  were equivalent to the results we obtained in the moderate/severe SDB group. The inter-class correlations for time in N3 and REM across our SDB subjects were on average 20% better than results obtained in healthy subjects using a wireless forehead system . The algorithms in this study was also equivalent to a five-state, single channel method based on sensor sites at Cz and Pz . When comparing the accuracy of our automated scoring algorithm with the latter method, it is important to note that the sensitivity and positive predictive values presented in this study are based on one scorer per study, with multiple technicians used to score the data set. Conversely, Berthomier et al.  used two expert scorers and eliminated epochs in which raters disagree, an approach that can artificially inflated accuracy metrics by discarding epochs found difficult to score by human experts.
Because sleep staging is time consuming and somewhat subjective in nature, the auto-scoring can introduce a degree of consistency in the characterization of sleep architecture prior to visual inspection, a favorable approach when sleep EEG is used to evaluate outcomes such as those based on cognitive behavioral therapy [32, 33]. Given the automated algorithm can be applied to stage sleep in real time and there were limited differences in RDI values calculated with total sleep times derived from manual vs. automated scoring from two electrooculographic leads, the algorithm could also be used to assist technicians identify two-hours of sleep time needed to transition from diagnosis to therapy in continuous positive airway pressure split night studies. Real time sleep staging could also be beneficial during assessment of sleep stage related RDI differences in patients undergoing an oral appliance titration study or in partial sleep-deprivation studies to allow the attending technician to monitor sleep stages of interest.
The combination of a lightweight, easily applied data acquisition system with automated sleep staging algorithms holds promise for routine application of sleep profiling in the evaluation of many disease states that are currently difficult to identify. As highlighted in the introduction, sleep disturbances are among the hallmark features of many psychiatric and neurological disorders and there is a need for objective biomarkers to confirm diagnosis, to identify sub-types and ultimately to ensure treatment efficacy. For example, periods of wake that serve as biomarkers of insomnia are distinct from the abnormally long and dense REM cycles for major depression. Differentiating depression from insomnia would be an important first step before introduction of a pharmacological or other intervention. In the case of depression, the normalization of the REM patterns, as seen in response to the Selective Serotonin Reuptake Inhibitors (SSRI), could be used as an indicator of treatment efficacy [34–36]. In contrast, sleep disturbances in PTSD and are often resistant to treatment with SSRIs [20, 37]. Although there is substantial symptom overlap between depression and PTSD (e.g. depressed mood, anhedonia, social withdrawal, and decreased concentration), PTSD is characterized by disturbed sleep continuity with decreased total sleep time, poor sleep efficiency, minimal slow wave sleep and long but fragmented REM. More recently it has been suggested that fragmented REM sleep, characteristic of many patients with PTSD, may disrupt the process of encoding traumatic memories that normally allows for their recollection without the emotional charge and associated sympathetic arousal [20, 38, 39]. Drugs such as prazosin or clonidine, which act antagonistically at noradrenergic receptors, have recently been shown to improve sleep and relieve symptoms in some PTSD patients specifically reducing the number and severity of nightmares [40, 41]. Thus, sleep profiling could be used to differentiate insomnia from depression or to indentify trauma victims at risk of developing PTSD and to improve the selection of therapy. Follow-up studies could evaluate the efficacy of prescribed pharmacological treatments in those diagnosed with psychiatric conditions.
The level of agreement between automated sleep staging and visual scoring across all sleep stages other than N1 is approximately the same as the inter-rater agreement between scorers. Decreased auto-staging accuracy corresponds with increased sleep disordered breathing severity. These results suggest that the automated sleep staging algorithm may prove useful in evaluating sleep architecture patterns in patients with chronic diseases. Further investigation is required to determine if biomarkers derived from the automated staging will assist in the diagnosis and treatment of chronic diseases.
Development of the sleep staging algorithm was supported in part by the Defense Advanced Research Projects Agency (DARPA) under SBIR grant W31P4Q-08-C-0123) and the National Institute of Health (SBIR 5R44NL68463-5). The views, opinions, and/or findings contained are those of the authors and should not be interpreted as representing the official views or policies, either expressed or implied, of the Defense Advanced Research Projects Agency or the Department of Defense. The authors are grateful to Dr. InduAyappa, Dr. David Rapoport and their staff at the New York University School of Medicine Sleep Disorders Center who recruited the subjects and acquired the reported data.
- Cappuccio FP, D’Elia L, Strazzullo P, Miller MA: Sleep duration and all-cause mortality: a systematic review and meta-analysis of prospective studies. Sleep 2010,33(5):585–592.PubMedGoogle Scholar
- Gangwisch JE, Heymsfield SB, Boden-Albala B, Buijs RM, Kreier F, Pickering TG, Rundle AG, Zammit GK, Malaspina D: Sleep duration as a risk factor for diabetes incidence in a large US sample. Sleep 2007,30(12):1667–1673.PubMedGoogle Scholar
- Tasali E, Laproult R, Ehrmann DA, Van Cauter E: Slow-wave sleep and the risk of type 2 diabetes in humans. PNAS 2008,105(3):1044–1049.PubMedView ArticleGoogle Scholar
- Knutson KL, Ryden AM, Mander BA, Van Cauter E: Role of sleep duration and quality in the risk and severity of type 2 diabetes mellitus. Arch Intern Med 2006,166(16):1768–1774.PubMedView ArticleGoogle Scholar
- Knutson KL, Van Cauter E: Association between sleep loss and increased risk of obesity and diabetes. Ann NY AcadSci 2008, 1129:287–304.View ArticleGoogle Scholar
- Spiegel K, Knutson K, Leproult R, Tasali E, Van Cauter E: Sleep loss: a novel risk factor for insulin resistance and Type 2 diabetes. J ApplPhysiol 2005,99(5):2008–2019.View ArticleGoogle Scholar
- Leproult R, Van Cauter E: Role of sleep and sleep loss in hormonal release and metabolism. EndocrDev 2010, 17:11–21.Google Scholar
- Theorell-Haglöw J, Berne C, Janson C, Sahlin C, Lindberg E: Associations between short sleep duration and central obesity in women. Sleep 2010,33(5):593–598.PubMedGoogle Scholar
- Benca RM, Obermeyer WH, Thisted RA, Gillin JC: Sleep and psychiatric disorders: a meta-analysis. Arch Gen Psychiatry 1992,49(8):651–668.PubMedView ArticleGoogle Scholar
- Benca RM: Diagnosis and treatment of chronic insomnia: a review. PsychiatrServ 2005,56(3):332–343.Google Scholar
- Ohayon MM: Insomnia: a ticking clock for depression? J Psychiatr Res 2007,41(11):893–894.PubMedView ArticleGoogle Scholar
- Ohayon MM: Epidemiology of insomnia: what we know and what we still need to learn. Sleep Med Rev 2002,6(2):97–111.PubMedView ArticleGoogle Scholar
- Ohayon MM: Prevalence and correlates of nonrestorative sleep complaints. Arch Intern Med 2005,165(1):35–41.PubMedView ArticleGoogle Scholar
- Ohayon MM: Nocturnal awakenings and comorbid disorders in the American general population. J Psychiatric Res 2008,43(1):48–54.View ArticleGoogle Scholar
- Fung MM, Peters K, Redline S, Ziegler MG, Ancoli-Israel S, Barrett-Connor E, Stone KL: Decreased slow wave sleep increases risk of developing hypertension in elderly men. Hypertension 2011,58(4):596–603.PubMedView ArticleGoogle Scholar
- Pallayova M, Donic V, Gresova S, Peregrim I, Tomori Z: Do differences in sleep architecture exist between persons with type 2 diabetes and nondiabetic controls? J Diabetes SciTechnol 2010,4(2):344–352.Google Scholar
- Rao MN, Blackwell T, Redline S, Stefanick ML, Ancoli-Israel S, Stone KL: Association between sleep architecture and measures of body composition. Sleep 2009,32(4):483–490.PubMedGoogle Scholar
- Pillai V, Kalmbach DA, Ciesla JA: A meta-analysis of electroencephalographic sleep in depression: evidence for genetic biomarkers. Biol Psychiatry 2011,70(10):912–919.PubMedView ArticleGoogle Scholar
- Postuma RB, Bertrand JA, Montplaisir J, Diesjardins C, Vendette M, Rios Romenets S, Panisset M, Gagnon JF: Rapid eye movement sleep behavior disorder and risk of dementia in Parkinson’s disease: A prospective study. MovDisord 2012. Feb epubGoogle Scholar
- Germain A, Buysse DJ, Nofzinger E: Sleep-specific mechanisms underlying posttraumatic stress disorder: integrative review and neurobiological hypotheses. Sleep Med Rev 2008,12(3):185–195.PubMedView ArticleGoogle Scholar
- Moser D, Anderer P, Gruber G, Parapatics S, Loretz E, Kloesch G, Heller E, Schmidt A, Danker-Hopfe H, Saletu B, Zeithofer J, Dorffner G: Sleep classification according to AASM and Rechtschaffen& Kales: effects on sleep scoring parameters. Sleep 2009,32(2):139–149.PubMedGoogle Scholar
- Norman RG, Pal I, Stewart C, Walsleben JA, Rapoport DM: Interobserver agreement among sleep scores from different centers in a large dataset. Sleep 2000,23(7):901–908.PubMedGoogle Scholar
- Danker-Hopfe H, Kunz D, Gruber G, Klosch G, Lorenzo JL, Himanen SL, Kemp B, Penzel T, Roschke J, Dorn H, Schlogl A, Trenker E, Dorffner G: Inter-rater reliability between scorers from eight European sleep laboratories in subjects with different sleep disorders. J Sleep Res 2004, 13:63–69.PubMedView ArticleGoogle Scholar
- Whitney CW, Gottlieb DJ, Redline S, Norman R, Dodge RR, Shahar E, Surovec S, Nieto FJ: Reliability of scoring respiratory disturbance indices and sleep staging. Sleep 1998,21(7):749–757.PubMedGoogle Scholar
- Virkkala J, Hasan J, Varri A, Himanen SL, Muller K: Automatic sleep stage classification using two channel electro-oculography. J Neuro Methods 2007, 166:109–115.View ArticleGoogle Scholar
- Shambroom JR, Fabregas S, Johnstone J: Validation of an automated wireless system to monitor sleep in healthy adults. J Sleep Res 2011. epubGoogle Scholar
- Berthomier C, Drouot X, Herman-Stoica M, Berthomier P, Prado J, Bokar-Thire D, Benoit O, Mattout J, D’ Ortho MP: Automated analysis of single-channel sleep EEG: validation in healthy individuals. Sleep 2007,30(11):1587–1595.PubMedGoogle Scholar
- Popovic D, Khoo M, Westbrook PR: Automatic detection of slow-wave sleep from a single forehead EEG channel. Euro J Sleep Res acceptedGoogle Scholar
- Rechtschafen A, Kales A: A manual of standardized terminology, techniques, and scoring system for sleep states of human subjects. U.S. Government Printing Office, Washington, DC; 1968.Google Scholar
- American Academy of Sleep Medicine Task Force: Sleep-related breathing disorders in adults: recommendations for syndrome definition and measurement techniques in clinical research. Sleep 1999, 22:667–689.Google Scholar
- Iber C, Ancoli-Israel S, Chesson A, Quan SF: he AASM Manual for the Scoring of Sleep and Associated Events: Rules, Terminology and Technical Specifications. American Academy of Sleep Medicine, Westchester; 2007.Google Scholar
- Krystal AD, Edinger JD: Sleep EEG predictors and correlates of the response to cognitive behavioral therapy for insomnia. Sleep 2010,33(5):669–677.PubMedGoogle Scholar
- Ulmer CS, Edinger JD, Calhoun PS: A multi-component cognitive-behavioral intervention for sleep disturbance in veterans with PTSD: a pilot study. J Clin Sleep Med 2011,7(1):57–68.PubMedGoogle Scholar
- Habukawa M, Uchimura N, Maeda M, Kotorii N, Maeda H: Sleep findings in young adult patients with posttraumatic stress disorder. Biol Psychiatry 2007,62(10):1179–1182.PubMedView ArticleGoogle Scholar
- Kobayashi I, Boarts JM, Delahanty DL: Polysomnographically measured sleep abnormalities in PTSD: a meta-analytic review. Psychophysiology 2007,44(4):660–669.PubMedView ArticleGoogle Scholar
- Babson KA, Feldner MT: Temporal relations between sleep problems and both traumatic event exposure and PTSD: a critical review of the empirical literature. J Anxiety Disord 2010,24(1):1–15.PubMedView ArticleGoogle Scholar
- Spoormaker VI, Montgomery P: Disturbed sleep in post-traumatic stress disorder: secondary symptom or core feature? Sleep Med Rev 2008,12(3):169–184.PubMedView ArticleGoogle Scholar
- Germain A: Sleep disturbances in posttraumatic stress disorder. Psychiatr Ann 2009,39(6):335–342.View ArticleGoogle Scholar
- Walker MP, Van Der Helm E: Overnight therapy? The role of sleep in emotional brain processing. Psychol Bull 2009,135(5):731–748.PubMedView ArticleGoogle Scholar
- Miller LJ: Prazosin for the treatment of posttraumatic stress disorder sleep disturbances. Pharmacotherapy 2008,28(5):656–666.PubMedView ArticleGoogle Scholar
- Taylor FB, Martin P, Thompson C, Williams J, Mellman TA, Gross C, Peskind ER, Raskind MA: Prazosin effects on objective sleep measures and clinical symptoms in civilian trauma posttraumatic stress disorder: a placebo-controlled study. Biol Psychiatry 2008,63(6):629–632.PubMedView ArticleGoogle Scholar
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.