Sleep in remitted bipolar disorder: A naturalistic case-control study using actigraphy

Abstract

Introduction
Findings from actigraphic studies suggesting that sleep and circadian rhythms are disrupted in bipolar disorder (BD) patients have been undermined by methodological heterogeneity and the failure to adequately address potential confounders.

Method
Twenty-six euthymic BD cases and 29 healthy controls (HC), recruited from University Paris-Est and matched for age and gender, were compared on subjective (Pittsburgh Sleep Questionnaire Inventory; PQSI) and objective (mean scores and variability in actigraphy) measures of sleep as recorded by over 21 consecutive days.

Results
Multivariate generalized linear modelling (GLM) revealed significant differences between BD cases and HC for five PSQI items (total score and four subscales), four actigraphy variables (mean scores) and five actigraphy variability measures. Backward stepwise linear regression (BSLR) indicated that a combination of four variables (mean sleep duration, mean sleep latency, variability of the fragmentation index over 21 days, and mean score on PSQI daytime dysfunction sub-scale) correctly classified 89% of study participants as cases or controls (Chi-square=39.81; df=6; p=0.001).

Limitations
The sample size (although larger than most actigraphy studies) and incomplete matching of cases and controls may have influenced our findings. It was not possible to control for potential effects of psychotropic medication or differences in employment status between groups.

Conclusions
When potential confounders of sleep and circadian profiles are adequately taken into account (particularly age, gender, daytime sleepiness, mood symptoms, body mass index, and risk of sleep apnoea), a selected subset of quantitative (mean scores) and qualitative (variability) features differentiated euthymic BD cases from HC.

Full text

Research report
Sleep in remitted bipolar disorder: A naturalistic case-control
study using actigraphy
Pierre Alexis Geoffroy
a,c,d,e,
n
, Carole Boudebesse
a,c,e
, Frank Bellivier
e,f,g
, Mohamed Lajnef
a
,
Chantal Henry
a,b,c,e
, Marion Leboyer
a,b,c,e
, Jan Scott
h,i
, Bruno Etain
a,c,e
a
INSERM, U955, Psychiatrie génétique, Créteil 94000, France
b
Université Paris Est, Faculté de médecine, Créteil 94000, France
c
AP-HP, Hôpital H. Mondor –A. Chenevier, Pôle de Psychiatrie, Créteil 94000, France
d
Pôle de psychiatrie, Université Lille Nord de France, CHRU de Lille, F-59000 Lille, France
e
Fondation FondaMental, Créteil 94000, France
f
AP-HP, GH Saint-Louis, Lariboisière, Fernand Widal, Pôle Neurosciences, Paris, France
g
Université Paris-7 Paris-Diderot, UFR de Médecine, Paris, France
h
Academic Psychiatry, Institute of Neuroscience, Newcastle University, UK
i
Centre for Affective Disorders, Institute of Psychiatry, London, UK
article info
Article history:
Received 1 November 2013
Received in revised form
15 January 2014
Accepted 16 January 2014
Available online 30 January 2014
Keywords:
Bipolar disorder
Actigraphy
Euthymia
Remission
Sleep
Circadian rhythms
abstract
Introduction: Findings from actigraphic studies suggesting that sleep and circadian rhythms are
disrupted in bipolar disorder (BD) patients have been undermined by methodological heterogeneity
and the failure to adequately address potential confounders.
Method: Twenty-six euthymic BD cases and 29 healthy controls (HC), recruited from University Paris-Est
and matched for age and gender, were compared on subjective (Pittsburgh Sleep Questionnaire
Inventory; PQSI) and objective (mean scores and variability in actigraphy) measures of sleep as recorded
by over 21 consecutive days.
Results: Multivariate generalized linear modelling (GLM) revealed significant differences between BD
cases and HC for five PSQI items (total score and four subscales), four actigraphy variables (mean scores)
and five actigraphy variability measures. Backward stepwise linear regression (BSLR) indicated that a
combination of four variables (mean sleep duration, mean sleep latency, variability of the fragmentation
index over 21 days, and mean score on PSQI daytime dysfunction sub-scale) correctly classified 89% of
study participants as cases or controls (Chi-square ¼39.81; df¼6; p¼0.0 01).
Limitations: The sample size (although larger than most actigraphy studies) and incomplete matching of
cases and controls may have influenced our findings. It was not possible to control for potential effects of
psychotropic medication or differences in employment status between groups.
Conclusions: When potential confounders of sleep and circadian profiles are adequately taken into
account (particularly age, gender, daytime sleepiness, mood symptoms, body mass index, and risk of
sleep apnoea), a selected subset of quantitative (mean scores) and qualitative (variability) features
differentiated euthymic BD cases from HC.
&2014 Elsevier B.V. All rights reserved.
1. Introduction
Bipolar disorder (BD) is a severe mental disorder and the world-
wide prevalence of the BD spectrum is 1–4%. Although researchers
agree that BD is multifactorial with inherited, genetic and environ-
mental risk factors, the patho-physiological determinants are
unknown (Geoffroy et al., 2013). Recent studies of circadian genes
and phenotypes indicate that sleep/activity dysregulation is a plausible
model for BD episode recurrence (McClung, 2013). Abnormalities in
sleep and circadian biomarkers during the euthymic phase have been
identified, revealing that this dysregulation is not purely an epiphe-
nomenon of acute illness episodes; indeed, increased disruption of the
sleep homoeostasis and circadian system are frequently associated
with relapse (Etainetal.,2011;Harvey,2008). Consequently, the
validation of sleep and circadian biomarkers of BD may help in the
development of both methods for earlier and more accurate detection
of unstable mental states, and of novel strategies to prevent relapse
(Frey et al., 2013).
Subjective reporting of sleep profilesisusefulinclinicalpractice,
although objective measures are increasingly utilized for research.
Contents lists available at ScienceDirect
journal homepage: www.elsevier.com/locate/jad
Journal of Affective Disorders
0165-0327/$- see front matter &2014 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.jad.2014.01.012
n
Corresponding author at: Pôle de Psychiatrie des Hopitaux Universitaires Henri
Mondor, Centre Expert Bipolaire, Hôpital Albert Chenevier, pavillon Hartman, 40
rue de Mesly, 94000 Créteil Cedex, France. Tel.: þ33 1 49 81 32 90;
fax: þ33 1 49 81 30 99.
E-mail address: pierre.a.geoffroy@gmail.com (P.A. Geoffroy).
Journal of Affective Disorders 158 (2014) 1–7

Actigraphy (recording the subject's sleep/wake cycles generally with
amobileportabledevice)isconsidered to be the most relevant non-
invasive technique for the measurement of sleep/wake irregularities
(Morgenthaler et al., 2007). Actigraphy has been used extensively in
the field of chronobiology, and has more recently been applied to
research in mood disorders, and in particular the characterization of
sleep homoeostasis and circadian rhythm abnormalities in BD
(Kaplan et al., 2012). We identified nine published actigraphic studies
of BD: although most suggest that, even in remission, BD cases were
more likely than ‘controls' to display a range of sleep abnormalities,
there was significant heterogeneity in the methodologies employed
(Gershon et al., 2012; Harvey et al., 2005; Jones et al., 2005; Kaplan
et al., 2012; Millar et al., 2004; Mullin et al., 2011; Ritter et al., 2012;
Salvatore et al., 2008; St-Amand et al., 2012). The five major
between-study differences were (1) the mean age of sample (e.g.
ranging from adolescents with early onset BD to older adults with
chronic,persistentBD);(2)theoverallsamplesize(thetotalnumber
of participants including cases and controls ranging from 27 to 68);
(3) the duration of actigraphic recording (some studies recorded data
for only 3 days with a median of 6 days for the nine studies); (4)
different approaches to ‘matching' cases and controls, with a number
of studies failing to match on demographic characteristics; and/or (5)
failure to control for confounding factors such as body mass index
(BMI) and/or sleep apnoea syndrome (SAS).
Most of these previous studies examined ‘quantitative’differences
(mean scores of each sleep measure for each group), and mainly
focused on four basic sleep variables (sleep duration, latency and
efficiency, and waking after sleep onset). However, it is increasingly
argued that differences in terms of quality or variability (which can
be represented by the standard deviations [SD] in the mean scores
for a particular measure extracted from actigraphic recordings) may
be more sensitive markers for distinguishing BD sleep patterns from
those of controls (Scott, 2011). Only three studies have published
data relevant to this approach (Gershon et al., 2012; Millar et al.,
2004; Mullin et al., 2011), but comparing the findings is difficult.
Indeed, the study among adolescents (Mullin et al., 2011), and the
study involving three days of actigraphic recording (Salvatore et al.,
2008) differed from the study in older adults over a longer period (28
days) of actigraphic monitoring (Gershon et al., 2012). Also, none of
these studies controlled for the possible effects of BMI or level of risk
for SAS on sleep profiles. Thus, published findings for the quality and
quantity of sleep in remitted BD may have been confounded by other
individual characteristics, including those known to affect sleep
directly, and by heterogeneity in sampling and/or in study design.
To overcome the limitations noted in previous studies, we report
an analysis of mean scores of subjective and objective measures, and
variability in objective measures, using methodological approaches
designed to avoid potential confounders. Previous studies examined
each reported sleep variable in independent analyses and none of
them adequately addressed potential confounders in their statistical
analytic strategy. We therefore tried to overcome some of these
previous weaknesses in design by examining which combination of
sleep measures best distinguishes cases from controls. Our hypoth-
esis was to identify and/or confirm reliable and robust circadian/
sleep markers differentiating euthymic BD cases from HC when
potential confounders of sleep and circadian differences are taken
into account (most notably, age, gender, daytime sleepiness, mood
symptoms, BMI and risk of sleep apnoea).
2. Materials and methods
2.1. Sample
With ethical approval from the Institutional review board,
written informed consent was obtained from 55 adults (BD ¼26;
HC¼29). The included cases were recruited from our university-
affiliated psychiatric clinic (University Paris Est) and the HC were
recruited from individuals attending the blood donor service at the
adjacent general hospital (Henri Mondor Hospital, Creteil).
The groups (BD and HC) were matched as closely as possible for
age and gender: perfect matching was not feasible, so we
increased the size of the HC group to ensure that we had a control
of the same age and/or a control of the same gender for each case
recruited. Individuals eligible for the study were included if,
during the preceding three months, they had (1) not experienced
any periods of severe sleep disruption due to somatic conditions
(e.g. organic insomnia/hypersomnia or sleep-wake disorders) and/
or any life event that may have altered their sleep patterns (e.g.
shift work, jet-lag, child birth, trauma, or somatic disease known
to be associated with sleep disturbances); (2) not been hospita-
lised or received a treatment that may disrupt sleep (e.g. for cases:
electro-convulsive therapy); (3) not been prescribed medication or
taken drugs that may disrupt sleep (e.g. sympathomimetic stimu-
lants, corticosteroids, thyroid hormones, antiarrhythmics, beta-
blockers, clonidine, diuretics, theophylline, and medications con-
taining alcohol or caffeine) and not changed either the dose or
type of psychotropic treatment.
Bipolar cases also had to fulfil the following inclusion criteria:
the DSM-IV criteria for BD according to the French version of the
Diagnostic Interview for Genetic Studies (DIGS) (Nurnberger et al.,
198 8); currently euthymic i.e. they scored o8 on both the Mon-
tgomery–Asberg Depression Rating Scale (MADRS; Montgomery
and Asberg, 1979) and the Young Manic Rating Scale (YMRS;
Young et al., 1978); and the International Society of Bipolar Disorder
task force criteria (Tohe n et a l., 20 09 ) for remission (i.e. they had not
experienced a BD episode in the prior three months). Patients
experiencing a BD relapse whilst participating in the study were
excluded.
Controls were assessed with the DIGS and the Family Interview
for Genetic Studies (Maxwell, 1992); individuals were excluded if
they had any personal history of DSM-IV disorders and family
history of schizophrenia, affective disorders and/or suicide attempts.
2.2. Assessment procedures
Cases and controls were first assessed for affective symptoms
using the MADRS and YMRS, and their BMI was calculated.
Participants then completed the Berlin Questionnaire (a measure
of the risk of SAS) (Netzer et al., 1999), and the Epworth Sleepiness
Scale (a measure daytime sleepiness) (Johns, 1991), to allow these
potential confounding factors to be taken into account.
2.2.1. Self-ratings of sleep profile
All participants recorded their subjective sleep quality using
the French version of the 19-item Pittsburgh Sleep Quality Index
(PSQI) (Blais et al., 1997). Here, we report the mean total score over
21 consecutive days and seven other specific measures of sleep
(with higher scores indicating lower quality or worse outcome):
sleep quality, latency, duration and efficiency, disturbances, use of
sleeping medication and daytime dysfunction (e.g. sleepiness and
enthusiasm).
2.2.2. Actigraphy
All participants were asked to wear an actigraph (AW-7
CamNtech) on the wrist of their non-dominant hand for the 21
consecutive days, the days for which they completed the PSQI self-
ratings. The AW-7 actigraph is an accelerometer that detects the
intensity and the amount of movement as a function of time. For
this study, data were sampled in one-minute epochs and partici-
pants were instructed to press the event-marker when they went
P.A. Geoffroy et al. / Journal of Affective Disorders 158 (2014) 1–72

to bed to go to sleep and when they got out of bed at the start of
the next day. They also completed a sleep diary for these 21 days.
Other requirements were kept to a minimum to try to minimise
intrusion into sleep/activity routines. Recording was made during
periods when the participants were not on vacation and not
involved in unusual activities.
Two experienced psychiatrists (PAG and CB) analysed the
actigraphic records of cases and controls using the sleep detection
algorithm provided by Actiwatch software (Actiwatch Activity &
Sleep Analysis Ltd CamNtech
s
7.28) and any incongruences
between the sleep diary and actigraphy rest periods were clarified
prior to statistical analyses (see Boudebesse et al., 2012 for details).
We selected the ten clinically interpretable actigraphic mea-
sures that have been most frequently reported in previous studies
(Gershon et al., 2012; Harvey et al., 2005; Jones et al., 2005; Kaplan
et al., 2012; Millar et al., 2004; Mullin et al., 2011; Ritter et al.,
2012; Salvatore et al., 2008; St-Amand et al., 2012). There were six
measures of basic sleep profile: time in bed, sleep duration, sleep
latency, wake after sleep onset (WASO), sleep efficiency, fragmen-
tation index (a measure of sleep continuity); two measures of
activity during sleep: mean activity in active periods of sleep,
mean sleep activity; and two measures of variability: inter-daily
stability and intra-daily variability. To facilitate comparison with
other studies that used different measures of variability, we also
estimated the variability of the basic sleep measures over 21 days
(using standard deviations (SD) of the mean scores) (see Millar
et al., 2004; Mullin et al., 2011 for the rationale and details of this
approach).
2.3. Statistical analysis
All analyses were performed using SPSS 19.0. We included
several covariates in the analysis: age and/or gender because
matching was incomplete, and also BMI, SAS, daytime sleepiness
(Epworth) and the presence of residual or low-grade mood
symptoms (as measured using the MADRS or YMRS) that can all
be associated with differences in sleep profile independently of
case-control group status (i.e. BD or HC). We undertook a power
calculation to assess the sample size required to detect a between-
group Effect Size (ES) of Z0.35 (classified as a medium ES
according to Cohen (1977) for self-rated and objective measures
of sleep whilst controlling for these potential confounders. We
thereby estimated that the sample size of 55 individuals had
480% power to detect differences at a significance level of 0.05
(and an ESZ0.35) in a multivariate analysis.
The analysis involved three stages. First, we checked the
normality of the distributions of variables and undertook log
transformation if required (none were needed). Second, we used
multivariate Generalized Linear Modelling (GLM) for the analysis
of sleep variables (self-ratings; actigraphy measures). GLM is a
flexible statistical model that incorporates normally distributed
dependent variables and categorical or continuous independent
variables (we also report Wilk's Lambda, to indicate the proportion
of generalized variance in the dependent variables explained by
the model and also to give the best estimate of overall power of
the analysis to differentiate between groups). We used bootstrap-
ping (i.e. resampling) to reduce biases introduced by any ‘outliers’
and to obtain the best estimates of summary statistics. Any
variables that were significantly different between groups (even
allowing for interactions between group and other categorical or
continuous variables) were then entered into a Backward Stepwise
Logistic Regression (BSLR) to assess the combination of these
selected variables that best classified individuals as cases or
controls. The proportion of cases correctly classified is reported
along with the odds ratios (OR) and 95% confidence intervals (95%
CI) for each variable included in the final optimal model.
3. Results
3.1. Socio-demographic and clinical characteristics
The sample is composed of 55 individuals (26 BD cases; 29 HC),
the mean age was about 54 years and more than half (30/55) were
female (Table 1). Cases fulfilled the criteria for BD subtype I
(n¼16) or BD subtype II (n¼10). One patient did not receive
medication, and nine patients received only one psychotropic
medication and 16 received Z2 medications. The psychotropic
treatments used were lithium (n¼12), anti-convulsants (n¼15),
atypical antipsychotics (n¼7), typical antipsychotics (n¼4), ben-
zodiazepines (n¼4), antidepressants (n¼5).
There were no statistically significant differences between the
groups for age, gender, BMI, level of risk of SAS, or daytime
sleepiness. Levels of mood symptoms were very low in both
groups, but the median score on the MADRS was slightly higher
for the BD group than controls (0.50 vs. 0.00); there were no
between-group differences on the YMRS.
3.2. Subjective and objective measures of sleep
Mean scores for a number of subjective and objective ratings of
sleep differed between cases and controls. There were statistically
significant differences between the groups for several PSQI and
actigraphy measures, even when taking into account gender, risk of
SAS, age, BMI, and scores on the Epworth, MADRS, and YMRS. When
considering the multivariate values for the model as a whole, there
was a significant difference between the patients with BD and HC
(Wilks' lambda, F¼90.89; df 26,16; p¼0.0001; statistical power, 0.93).
Several interactions between group and other variables were
identified; in particular, five PSQI items (Total score and scores for
PSQI subscales 1, 4, 5 and 7), four mean actigraphy measures (sleep
latency, duration, efficacy, and fragmentation index, inter-daily stabi-
lity) and five actigraphy variability measures (variability in sleep
duration, sleep efficiency, fragmentation index and time in bed)
differed between the groups (Tabl e 2). As measured by the PSQI,
patients with BD were more likely to report poorer sleep quality
(p¼0.001), poorer sleep efficiency (p¼0.001), more frequent sleep
disturbances (p¼0.001), and more daytime dysfunction (p¼0.001);
consequently they had poorer global sleep quality as measured by the
PSQI total score (p¼0.0001). As determined from mean actigraphy
scores, patients with BD were more likely to have longer sleep latency
Table 1
Baseline characteristics of bipolar disorder (BD) cases and healthy controls (HC).
Baseline BD (n¼26) HC (n¼29) p
characteristics Mean (7SD) Mean (7SD)
Age (years) 53.50 (711.49) 54.10 (79.11) 0.78
Montgomery–Asberg 1.85 (72.80) 0.48 (71.35) 0.02
Depression rating scale
Median score 0.50 0.00
Young mania rating Scale 0.65 (71.38) 0.14 (70.44) 0.16
Median score 0.00 0.00
Body mass index 26.74 (75.72) 26.64 (74.17) 0.68
Daytime sleepiness
b
8.85 (73.90) 6.86 (73.72) 0.12
n(%) n(%)
Sleep apnoea syndrome
a
0.09
High risk category 6 (23%) 2 (7%)
Low risk category 20 (77%) 27 (93%)
Gender distribution 0.13
Female 17 (65%) 13 (45%)
Male 9 (35%) 16 (55%)
a
Risk of sleep apnoea syndrome assessed using the Berlin Questionnaire.
b
Daytime sleepiness assessed using the Epworth Sleepiness Scale.
P.A. Geoffroy et al. / Journal of Affective Disorders 158 (2014) 1–73

(p¼0.007), longer sleep duration (p¼0.02), poorer sleep efficiency
(p¼0.04), higher fragmentation index (p¼0.05) and poorer inter-daily
stability (p¼0.03). For actigraphy variability over 21 days, patients
with BD were more likely to show more variable time in bed
(p¼0.005), sleep duration (p¼0.04), sleep efficiency (p¼0.03) and
fragmentation index (p¼0.03).
Table 2
Differences between sleep measures for cases (BD) and controls (HC).
Sleep measures BD (n¼26) HC (n¼29) Multivariate generalized linear model
a
Mean (7SD) Mean (7SD) Fp Observed power
PSQI measures
Total score 7.38 (73.49) 4.11 ( 72.13) 4.45 0.0 001 0.99
Sub-scale 1 1.19 0.71 3.64 0.001 0.99
Subjective sleep quality (70.69) ( 70.66)
Sub-scale 2 1.19 0.82 0.70 0.74 0.33
Sleep latency (70.80) ( 70.91)
Sub-scale 3 0.62 0.50 1.53 0.15 0.70
Sleep duration (70.94) ( 70.75)
Sub-scale 4 0.88 0.43 3.86 0.001 0.99
Habitual sleep efficiency (70.95) ( 70.69)
Sub-scale 5 1.58 1.18 3.67 0.001 0.99
Sleep disturbances (70.58) ( 70.48)
Sub-scale 6 0.69 0.14 1.08 0.18 0.68
Use of night sedation (71.09) (70.59)
Sub-scale 7 1.23 0.32 3.53 0.001 0.98
Daytime dysfunction (70.99) ( 70.48)
Mean actigraphy scores
Time in bed (min) 511.81 ( 750.28) 482.31 (749.36) 1.79 0.08 0.79
Sleep duration (min) 475.42 (764.5) 455.83 (753.98) 2.38 0.019 0.91
Sleep latency (min) 25.23 (733.65) 11.59 ( 77.98) 2.81 0.007 0.95
Wake after sleep onset (min) 57.88 (723.17) 52.62 ( 727.33) 0.57 0.85 0.27
Sleep efficiency (%) 81.54 (79.87) 84.90 (76.32) 2.13 0.036 0.86
Fragmentation index 32.04 (710.51) 28.69 (710.88) 2.01 0.05 0.84
Mean activity in active sleep periods
b
115.54 ( 737.67) 91.10 (724.39) 1.21 0.31 0.58
Mean overall sleep activity 19.35 ( 713.10) 12.69 ( 77.07) 1.69 0.11 0.76
Inter-daily stability 0.503 (70.14) 0.512 (70.12) 2.17 0.033 0.87
Intra-daily variability 0.82 ( 70.17) 0.80 ( 70.18) 0.90 0.56 0.43
Actigraphy measure variability over 21 days
Time in bed (min) 1.44 (70.7) 1.33 (70.5) 2.97 0.005 0.96
Sleep duration (min) 1.3 (70.6) 1.13 (70.4) 2.08 0.04 0.86
Sleep latency (min) 0.67 (71.2) 0.43 (70.7) 1.40 0.21 0.66
Wake after sleep onset (min) 0.49 (70.3) 0.36 ( 70.3) 1.80 0.08 0.79
Sleep efficiency (%) 8.77 (77.5) 6.13 ( 73.8) 2.26 0.026 0.89
Fragmentation index 12.52 (77.3) 9.28 (72.7) 2.16 0.033 0.96
a
Model controlled for age, gender, daytime sleepiness (Epworth), current mood symptoms (MADRS and YMRS), BMI and risk of sleep apnoea (Berlin).
b
Intensity of movement during periods of sleep when activity occurs.
Table 3
Backward Stepwise Logistic Regression (BSLR) showing (a) the best combination of variables for correctly classifying study participants as
cases or controls and (b) the overall classification rate by group.
Variable 95% C.I.
pOdds Ratio Lower Upper
(a) Variables included in the final BSLR model
Sleep duration 0.015 1.03 1.01 1.05
Sleep latency 0.012 1.24 1.05 1.47
Fragmentation index: mean variability 0.01 1.63 1.12 2.31
Daytime dysfunction (PSQI subscale-7) 0.005 8.35 2.50 18.11
Observed Predicted
Controls Cases % Correctly classified
(b) Classification table and summary statistics
Controls
a
25 3 89.3
Cases 3 23 88.5
% of all participants correctly classified 89%.
Chi-square 39.81 (df 6) p¼0.001.
a
One control participant was excluded from the analysis due to inadequate data.
P.A. Geoffroy et al. / Journal of Affective Disorders 158 (2014) 1–74

3.3. Regression analysis
We next used BSLR to determine the optimal combination of
variables that best classified participants as cases or controls.
When the variables identified by the GLM were entered into
a BSLR; a combination of four variables (mean sleep duration,
mean sleep latency, variability in fragmentation index over 21
days, and mean score on PSQI sub-scale 7: daytime dysfunction)
correctly classified 89% of the study participants as cases or
controls (Chi-square¼39.81; df ¼6; p¼0.001). The OR for these
variables ranged from 1.03 (sleep duration) to 8.35 (PSQI ‘daytime
dysfunction’sub-scale) (Table 3). However, whilst daytime dys-
function is the variable with the highest OR, on its own it classified
about 75% cases, so the other measures significantly contributed to
the improved classification obtained by the final model.
4. Discussion
We report a study of sleep abnormalities in remitted bipolar
(BD) patients compared with HC that also controlled for major
potential confounders, and replicate and extend previous findings.
In particular, objective and subjective assessments of sleep indi-
cate that, compared to controls, BD patients show longer sleep
duration, longer sleep latency, poorer sleep efficiency, higher
fragmentation index (with actigraphy) (Gershon et al., 2012;
Harvey et al., 2005; Jones et al., 2005; Kaplan et al., 2012; Millar
et al., 2004; Mullin et al., 2011; Ritter et al., 2012; Salvatore et al.,
2008; St-Amand et al., 2012), lower sleep quality, weaker sleep
efficiency, more sleep disturbances, and more daytime dysfunction
(with PSQI) (Rocha et al., 2013). The partial discrepancy between
PSQI (subjective measure) and actigraphy (objective measure)
findings has also been observed previously: patients with BD
underestimated their sleep latency and duration on questionnaire
assessments compared to actigraphy examination. This may arise
as a consequence of misperceptions in patients with BD regarding
their sleep quality (Gershon et al., 2012; Harvey et al., 2005; Millar
et al., 2004). Nevertheless, actigraphy demonstrated a high corre-
lation with polysomnography (PSG) measures in patients with BD
regarding sleep latency, sleep duration, fragmentation index and
sleep efficiency, which further validate our findings (Kaplan et al.,
2012).
We report several new findings for issues that have been
under-explored: patients with BD presented with more variability
for time in bed, sleep duration, sleep efficiency and fragmentation
index. Indeed, sleep duration, sleep efficiency and fragmentation
index are altered not only in quantity (mean) but also in variability
(SD) in BD patients. This indicates that analyses of actigraphic
markers should not be restricted to classical sleep markers
(latency, duration, efficiency and WASO), but should also
be extended to the variability of markers. Indeed, Millar et al.
(2004) were the first to explore the variability of markers in
different groups and showed that the means of none of the four
actigraphic measures differed between cases and controls, but
variability of two of the four measures did (sleep duration and
night waking time) with an effect size around 0.35. Similarly,
Gershon et al. (2012) confirmed more variability of both time in
bed and sleep duration for BD patients than controls (although the
latter did not survive their Bonferroni's correction) Millar et al.
(2004) also found more variability of night wake time among
remitted cases. Both Gershon et al. (2012) and our study
(0.4070.3 vs. 0.3670.3) replicate this finding. Further, we
observed significant variability of sleep efficiency and fragmenta-
tion index, which confirmed the greater variability of sleep
patterns and circadian rhythms in patients with remitted BD than
controls. Mullin et al. (2011), despite also assessing the variability
of actigraphic measures, did not find any significant differences;
however, that study involved recordings for only four nights,
which may have been too short to detect variability. Studies
involving sufficient duration of recording would be useful to
explore further the variability of actigraphic measures in BD
patients, and to examine trait-markers of BD.
The results of our regression analysis deserve comment. The
fact that 89.3% of BD and 88.5% of HC could be correctly classified
indicates that a combination of different methods of assessment
(i.e. PSQI, mean and SD actigraphy measures) may serve as a useful
marker of BD. Daytime dysfunction, sleep duration, sleep latency
and fragmentation index variability appeared, when combined, to
identify clear circadian biomarkers of remitted BD. Indeed, it may
be possible to define a biosignature of BD using these four
features. We partially replicate the findings of Millar et al. (2004)
whose best multivariate model involved a combination of one
actigraphic (variability of sleep duration) and two subjective sleep
variables (average sleep duration, and average onset latency).
These two studies indicate that individuals are best classified by
a combination of subjective and objective measures (quantity and
variability). The other comment is that it is noteworthy that –
although daytime dysfunction –was a sub-scale score of the PSQI,
it is debatable whether this is a specific measure of sleep pattern/
dysfunction. Indeed in BD cases it may be a measure of overall
effects of illness or possible effects of medication, not only of
daytime consequences of disturbed sleep.
Our study has some limitations. Controlling for psychotropic
medications is always difficult in bipolar disorder research. In this
study, the size of the case sub-group did not allow us to include
treatment data as a covariate in any ‘within group' sub-analyses.
However, the type and dose of psychotropic medications have
been found to be unrelated to actigraphic measures in patients
with remitted BD (Salvatore et al., 2008). Moreover, we believe
that on balance, it can be hypothesized that medication would be
as likely or more likely to improve the patients' quality of sleep as
to have a detrimental effect. We also attempted to control for
daytime sedation by using scores on the Epworth scale as a
covariate in the preliminary between group comparisons. As such,
the observed differences between cases and controls may have
been attenuated rather than exaggerated by such treatments.
Indeed, antipsychotics, valproate and lithium have been shown
to regulate sleep and circadian rhythms in patients with bipolar
disorder (Geddes and Miklowitz, 2013). Therefore, we believe that
it is unlikely that psychotropic drugs had a major influence. Our
sample size is small; however, it is of reasonable size for an
actigraphic study and larger than the median for the previous
studies cited (median¼40) (Gershon et al., 2012; Harvey et al.,
2005; Jones et al., 2005; Kaplan et al., 2012; Millar et al., 2004;
Mullin et al., 2011; Ritter et al., 2012; Salvatore et al., 2008; St-
Amand et al., 2012). We studied several markers of subjective and
objective sleep. Although GLM is an appropriate method for
assessing multiple, potentially inter-dependent measures, and
the power calculation (and estimated effect sizes) suggest we
have reduced the risk of type II error, it is clear that it would be
beneficial to use a larger sample for further studies. Nevertheless,
our study allowed controlling for confounding factors that were
insufficiently taken into account in previous studies; our greater
attention to potential confounding factors was an important
methodological strength relative to most previous studies.
We were thus able to confirm that patients with BD in
remission experienced substantial sleep disturbances that can be
measured both objectively and subjectively. This provides further
evidence of the involvement of sleep and circadian rhythm
abnormalities in the pathophysiology of BD, as proposed by, e.g.,
Harvey et al. (Harvey, 2008). Indeed, genetic variants of candidate
genes (mainly circadian genes) may predispose individuals to be
P.A. Geoffroy et al. / Journal of Affective Disorders 158 (2014) 1–75

relatively less able to adapt their circadian rhythms appropriately
to their environment, and therefore to greater vulnerability to
sleep disturbances. Since circadian and neurotransmission systems
are highly connected, circadian and/or sleep-related abnormalities
may affect the functioning of the dopamine and serotonin circui-
try, which in turn may affect mood regulation. These chronic sleep
and circadian rhythm disturbances (observable in acute episodes
and in remission) may thereby contribute to the more general issues
persistently encountered bipolar patients such as poor quality of
cognitive functioning, residual emotional hyper-reactivity or vulner-
ability to metabolic disturbances (Boland et al., 2012; Boudebesse
and Henry, 2012; McClung, 2013; Soreca et al., 2012).
The findings we report have several important clinical implica-
tions. Sleep and circadian rhythm disturbances in remitted BD
might be treatable by manipulating the circadian system using
chronobiological medication (e.g. melatonin or melatonin analo-
gues) or sleep focused psychological interventions (Kaplan and
Harvey, 2013; Livianos et al., 2012; Pacchierotti et al., 2001;
St-Amand et al., 2012). Attention to rhythm abnormalities is of
particular interest because biological rhythm dysfunctions have
been found to be associated to a wide range of sources of poor
functioning and are the earliest markers of impending mood
relapses (Giglio et al., 2010). Thus, better identification of these
abnormalities in clinical practice might facilitate the prevention of
the evolution of prodromes of mood relapses, and also help
improve functioning in inter-episodic BD. It may allow a more
complete approach including combinations of treatments acting
on several dimensions including sleep or circadian disturbances
(Geoffroy et al., 2012). These findings more generally indicate
a new area for research both into possible therapeutic targets
and to improve our understanding of the circadian pathophysiol-
ogy of BD.
5. Conclusion
This 21-day case-control study demonstrates the presence of
subjective and objective disturbances of sleep and activity markers
in patients with BD during remission. Disturbances identified by
actigraphy are both quantitative (mean scores) and qualitative
(variability). We further confirm the greater variability in sleep
markers in remitted patients with BD. These findings indicate that
future studies should examine both the mean scores and the
variability over extended periods of time. In addition, we show
that a combination of subjective and objective measures (quantity
and variability) may be a better circadian biosignature of BD than
any single measure on its own.
Role of funding source
This work was supported by AP-HP (Assistance Publique des Hôpitaux de Paris)
(Promotor of the study and PHRC AOR11096 PI : BE); the Fondation FondaMental (RTRS
Santé Mentale) (for funding the data management of database) and the Investissements
d'Avenir program managed by the ANR under reference ANR-11-IDEX-0004–02. This
work was also supported by Laboratoires Servier (funding for the purchase actiwatches
and for sequence analyses). Laboratoires Servier had no responsibility for the design of
the protocol, or for the acquisition, analysis or interpretation of data.
Conflict of interest
PA Geoffroy has received a prize from Bayer for being Laureate of the Medical
University of Lille.
C. Boudebesse has received a research bursary from Laboratoires Servier; a
prize for her thesis in medicine from Sanofi-Aventis; and honoraria from Ostuka as
an independent symposium speaker.
B. Etain and F. Bellivier have received honoraria and financial compensation as
independent symposium speakers from Sanofi-Aventis, Lundbeck, AstraZeneca, Eli
Lilly, Bristol-Myers Squibb and Servier.
C. Henry has received honoraria and financial compensation as independent
symposium speakers from Sanofi-Aventis, Lundbeck, AstraZeneca, Eli Lilly, Bristol-
Myers Squibb.
J. Scott has received funding to attend national and international conferences,
financial compensation as an independent symposium speaker for talks on early
onset in BD and psychosocial aspects of BD, and advisory board fees from
AstraZeneca, BMS-Otsuka, Eli Lilly, GSK, Jansen-Cilag, Lundbeck, Sanofi-Aventis
and Servier.
M. Leboyer has received honoraria and financial compensation as an indepen-
dent symposium speaker from AstraZeneca and Servier.
Acknowledgements
We thank the patients with BD and controls who agreed to participate in this
study. We thank the staff at the inclusion sites in Paris-Créteil (especially E. Abadie
for her participation in the clinical assessment). We are grateful to the Clinical
Investigation Centre (O. Montagne and P. Le Corvoisier). We thank J.R. Richard,
L. Ouvrard and N. Bord for their assistance.
References
Blais, F.C., Gendron, L., Mimeault, V., Morin, C.M., 1997. Evaluation of insomnia:
validity of 3 questionnaires. Encephale 23, 447–453.
Boland, E.M., Bender, R.E., Alloy, L.B., Conner, B.T., Labelle, D.R., Abramson, L.Y.,
2012. Life events and social rhythms in bipolar spectrum disorders: an
examination of social rhythm sensitivity. J. Affect. Disord. 39, 264–272 .
Boudebesse, C., Henry, C., 2012. [Emotional hyper-reactivity and sleep disturbances
in remitted patients with bipolar disorders]. Encephale 38 (4), S173–S178.
Boudebesse, C., Leboyer, M., Begley, A., Wood, A., Miewald, J., Hall, M., Frank, E.,
Kupfer, D., Germain, A., 2012. Comparison of five actigraphy scoring methods
with bipolar disorder. Behav. Sleep. Med. 11, 275–282.
Cohen, J., 1977. Statistical Power Analysis for the Behavioral Sciences, 2nd edition
Academic Press, New-York, NY.
Etain, B., Milhiet, V., Bellivier, F., Leboyer, M., 2011. Genetics of circadian rhythms
and mood spectrum disorders. Eur. Neuropsychopharmacol. 21 (4), S676–S682.
Frey, B.N., Andreazza, A.C., Houenou, J., Jamain, S., Goldstein, B.I., Frye, M.A.,
Leboyer, M., Berk, M., Malhi, G.S., Lopez-Jaramillo, C., Taylor, V.H., Dodd, S.,
Frangou, S., Hall, G.B., Fernandes, B.S., Kauer-Sant'Anna, M., Yatham, L.N.,
Kapczinski, F., Young, L.T., 2013. Biomarkers in bipolar disorder: a positional
paper from the International Society for Bipolar Disorders Biomarkers Task
Force. Aust. N. Z. J. Psychiatry 47, 321–332.
Geddes, J.R., Miklowitz, D.J., 2013. Treatment of bipolar disorder. Lancet 381,
1672 –168 2.
Geoffroy, P.A., Etain, B., Henry, C., Bellivier, F., 2012. Combination therapy for manic
phases: a critical review of a common practice. CNS Neurosci. Ther. 18,
957–964.
Geoffroy, P.A., Etain, B., Scott, J., Henry, C., Jamain, S., Leboyer, M., Bellivier, F., 2013.
Reconsideration of bipolar disorder as a developmental disorder: importance of
the time of onset. J. Physiol. Paris 07, 278–285.
Gershon, A., Thompson, W.K., Eidelman, P., McGlinchey, E.L., Kaplan, K.A., Harvey, A.G.,
2012.Restlesspillow,ruffled mind: sleep and affect coupling in interepisode
bipolar disorder. J. Abnorm. Psychol. 121, 863–873.
Giglio, L.M., Magalhães, P.V.S., Kapczinski, N.S., Walz, J.C., Kapczinski, F., 2010.
Functional impact of biological rhythm disturbance in bipolar disorder. J.
Psychiatr. Res. 44, 220–223.
Harvey, A.G., 2008. Sleep and circadian rhythms in bipolar disorder: seeking
synchrony, harmony, and regulation. Am. J. Psychiatry 165, 820–829.
Harvey, A.G., Schmidt, D.A., Scarnà, A., Semler, C.N., Goodwin, G.M., 2005. Sleep-
related functioning in euthymic patients with bipolar disorder, patients with
insomnia, and subjects without sleep problems. Am. J. Psychiatry 162, 50–57.
Johns, M.W., 1991. A new method for measuring daytime sleepiness: the Epworth
sleepiness scale. Sleep 14, 540–545.
Jones, S.H., Hare, D.J., Evershed, K., 2005. Actigraphic assessment of circadian
activity and sleep patterns in bipolar disorder. Bipolar Disord. 7, 176–186 .
Kaplan, K.A., Harvey, A.G., 2013. Behavioral treatment of insomnia in bipolar
disorder. Am. J. Psychiatry 170, 716–720.
Kaplan, K.A., Talbot, L.S., Gruber, J., Harvey, A.G., 2012. Evaluating sleep in bipolar
disorder: comparison between actigraphy, polysomnography, and sleep diary.
Bipolar Disord. 14, 870–879.
Livianos, L., Sierra, P., Arques, S., García, A., Rojo, L., 2012. Is melatonin an adjunctive
stabilizer? Psychiatry Clin. Neurosci. 66, 82–83.
Maxwell, M., 1992. Family Interview For Genetic Studies. Clinical Neurogenetics
Branch, Intramural Research Program NIMH.
McClung, C.A., 2013. How might circadian rhythms control mood? Let me count the
ways…. Biol. Psychiatry 74, 242–249.
Millar, A., Espie, C.A., Scott, J., 2004. The sleep of remitted bipolar outpatients: a
controlled naturalistic study using actigraphy. J. Affect. Disord. 80, 145–153.
Montgomery, S.A., Asberg, M., 1979. A new depression scale designed to be
sensitive to change. Br. J. Psychiatry: J. Ment. Sci. 134, 382–389.
Morgenthaler, T., Alessi, C., Friedman, L., Owens, J., Kapur, V., Boehlecke, B., Brown,
T., Chesson , A., Coleman, J., Lee-Chiong, T., Pancer, J., Swick, T.J., 2007. Practice
parameters for the use of actigraphy in the assessment of sleep and sleep
disorders: an update for 2007. Sleep 30, 519–529.
P.A. Geoffroy et al. / Journal of Affective Disorders 158 (2014) 1–76

Mullin, B.C., Harvey, A.G., Hinshaw, S.P., 2011. A preliminary study of sleep in
adolescents with bipolar disorder, ADHD, and non-patient controls. Bipolar
Disord. 13, 425–432.
Netzer, N.C., Stoohs, R.A., Netzer, C.M., Clark, K., Strohl, K.P., 1999. Using the Berlin
Questionnaire to identify patients at risk for the sleep apnea syndrome. Ann.
Intern. Med. 131, 485–491.
Nurnberger , J.I., Berrettini, W., Tamarkin, L., Hamovit, J., Norton, J., Gershon, E.,
1988. Supersensitivity to melatonin suppression by light in young people at
high risk for affective disorder. A preliminary report. Neuropsychopharmacol.
Off. Publ. Am. Coll. Neuropsychopharmacol. 1, 217–223.
Pacchierotti, C., Iapichino, S., Bossini, L., Pieraccini, F., Castrogiovanni, P., 2001.
Melatonin in psychiatric disorders: a review on the melatonin involvement in
psychiatry. Front. Neuroendocrinol. 22, 18–32.
Ritter, P.S., Marx, C., Lewtschenko, N., Pfeiffer, S., Leopold, K., Bauer, M., Pfennig, A.,
2012. The characteristics of sleep in patients with manifest bipolar disorder,
subjects at high risk of developing the disease and healthy controls. J. Neural
Transm. 119, 1173–1184 .
Rocha, P.M.B., Neves, F.S., Corrêa, H., 2013. Significant sleep disturbances in
euthymic bipolar patients. Compr. Psychiatry 54, 1003–10 08.
Salvatore, P., Ghidini, S., Zita, G., De Panfilis, C., Lambertino, S., Maggini, C.,
Baldessarini, R.J., 2008. Circadian activity rhythm abnormalities in ill and
recovered bipolar I disorder patients. Bipolar Disord. 10, 256–265.
Scott, J., 2011. Clinical parameters of circadian rhythms in affective disorders. Eur.
Neuropsychopharmacol. J. Eur. Coll. Neuropsychopharmacol. 21 (4), S671–S675.
Soreca, I., Wallace, M.L., Frank, E., Hasler, B.P., Levenson, J.C., Kupfer, D.J., 2012. Sleep
duration is associated with dyslipidemia in patients with bipolar disorder in
clinical remission. J. Affect. Disord. 141, 484–487.
St-Amand, J., Provencher, M.D., Bélanger, L., Morin, C.M., 2012. Sleep disturbances
in bipolar disorder during remission. J. Affect. Disord. 146, 112–119 .
Tohen, M., Frank, E., Bowden, C.L., Colom, F., Ghaemi, S.N., Yatham, L.N., Malhi, G.S.,
Calabrese, J.R., Nolen, W.A., Vieta, E., Kapczinski, F., Goodwin, G.M., Suppes, T.,
Sachs, G.S., Chengappa, K.R., Grunze, H., Mitchell, P.B., Kanba, S., Berk, M., 2009.
The International Society for Bipolar Disorders (ISBD) task force report on the
nomenclature of course and outcome in bipolar disorders. Bipolar Disord. 11,
453–473.
Young, R.C., Biggs, J.T., Ziegler, V.E., Meyer, D.A., 1978. A rating scale for mania:
reliability, validity and sensitivity. Br. J. Psychiatry: J. Ment. Sci. 133, 429–435.
P.A. Geoffroy et al. / Journal of Affective Disorders 158 (2014) 1–77