Potential conflict of interest: Nothing to report.
A significant proportion of patients with primary biliary cirrhosis (PBC) suffer from severe fatigue. The aim of this study was to characterize patterns of daytime sleep in patients with PBC (using both objective and subjective assessment approaches) and to study the association between sleep abnormality and fatigue severity. Fatigue severity was assessed in 48 female subjects with PBC (using a disease-specific quality of life instrument (the PBC-40) and a generic fatigue measure (Fatigue Impact Scale [FIS]) as well as 48 case-matched normal controls. All participants also completed the Pittsburgh Sleep Quality Index (PSQI) and the Epworth Sleepiness Scale (ESS, which assesses daytime hypersomnolence). Objective sleep assessment was performed using accelerometry over 7 days. Global sleep quality assessed by the PSQI was significantly lower in the PBC group compared to controls (P < .0001). ESS scores were significantly higher in patients with PBC than controls (P = .0001), suggesting significantly greater daytime somnolence in the patients with PBC. Objective sleep assessment confirmed that subjects with PBC were sleeping on average almost twice as long as controls during the daytime. Both degree of daytime somnolence (ESS) and actual daytime sleep activity (accelerometry) correlated strongly with fatigue severity in the patient group (r2 = 0.5, P < .0001 and r2 = 0.2, P < .01, respectively). In conclusion, Sleep abnormality, in the form of excessive daytime somnolence, is present in a significant proportion of patients with PBC, with the degree of daytime somnolence correlating strongly with the degree of fatigue. Existing agents effective at reducing daytime somnolence (such as modafinil) hold potential for the treatment of fatigue in PBC. (HEPATOLOGY 2006;44:91–98.)
Patients with the autoimmune liver disease primary biliary cirrhosis (PBC) experience a wide range of symptoms that can impact significantly on their quality of life (QOL).1, 2 The most prominent of these symptoms in many patient populations with PBC (including that in the UK) is fatigue.3–6 Despite its prevalence (up to 50% of patients are affected in some populations) and impact, little is known about the pathogenesis of fatigue in PBC, and no effective therapies are currently available. Numerous studies have failed to demonstrate a significant association between any conventional parameter of liver disease severity and the degree of fatigue experienced by patients with PBC.3, 6–8
One of the additional symptoms described most frequently by patients with PBC when participating in a recent comprehensive symptom impact study (performed to allow the derivation of a PBC-specific QOL measure [the PBC-40]) was abnormality in sleep, with patients particularly emphasizing excessive sleepiness during the daytime.1 This is of potential relevance for our understanding of the pathogenesis of fatigue in PBC as, in a number of non-hepatic conditions, fatigue has frequently been associated with abnormality in sleep patterns in general and daytime somnolence in particular. This has led to the suggestion that factors causing disruption in sleep rhythms may play an important role in the pathogenesis of fatigue.9, 10 Sleep abnormality has previously been documented in cases of PBC, with patients who self-reported fatigue showing the greatest levels of abnormality.3 The tools used to measure sleep in this study were not, however, optimal for the specific study of daytime somnolence, the aspect of sleep abnormality seemingly most strongly associated with fatigue. In the current study, therefore, we set out to build on these important earlier observations to study the specific issue of daytime sleep patterns and their associations with fatigue in PBC, using both subjective (the Epworth Sleepiness Scale) and objective (Actigraphy) measurement approaches for daytime sleep, and well-validated PBC-specific (PBC-40 Fatigue domain) and generic (Fatigue Impact Scale [FIS]) fatigue assessment tools.
PBC, primary biliary cirrhosis; QOL, quality of life; BMI, body mass index; FIS, Fatigue Impact Score; PSQI, Pittsburgh Sleep Quality Index; ESS: Epworth Sleepiness Scale; OSA, obstructive sleep apnea; AHI, apnea-hypopnea index.
Patients and Methods
Sleep and its relationship to fatigue were assessed in 48 female patients with PBC and 48 fully age-matched normal female control subjects (mean age, 62 ± 2 years for both subject groups). All subjects with PBC had definite or probable PBC using standard diagnostic criteria.5–8 None of the patients with PBC had clinically advanced disease or a clinical history of encephalopathy. All had normal liver synthetic function (bilirubin, albumin, and prothrombin time were within normal limits in all subjects). Clinical details of the patients with PBC participating in the study are outlined in Table 1. Normal control patients were recruited via notices within the hospital and through local voluntary organizations. Neither patients with PBC nor normal controls were questioned regarding sleep before enrollment, and none had previously undergone clinical investigation or management related to sleep abnormality. Subjects were excluded if they were taking therapeutic agents potentially able to modify sleep patterns or cause fatigue (including beta-blockers, opiate-based analgesics, benzodiazepines, other hypnotics, anti-histamines, and hypnotic, mood-altering or sedative herbal medications) or if they were suffering from other conditions associated with either fatigue or abnormality of sleep (including active thyroid disease, anemia, diabetes, renal impairment, and heart failure). Body mass index (BMI) was calculated from weight and height for all participants. The study was approved by the local research ethics committee, and all patients provided written informed consent.
Table 1. Clinical Details of Patients With PBC Participating in the Study
8.3 ± 3.4
41.8 ± 2.7
Prothrombin time (s)
12.1 ± 1.1
Alkaline phosphatase (U/L)
162 ± 101
Alanine transaminase (U/L)
42.2 ± 39.0
Advanced disease (III/IV) (% of those undergoing biopsy)
13.6 ± 1.2
2.4 ± 1.8
All subjects with PBC completed the PBC-40, a fully validated, PBC-specific, multi-domain QOL measure.1 The PBC-40 contains 40 questions in five domains (fatigue, itch, cognitive, social and emotional, and other symptoms [a domain relating to a number of PBC-related symptoms that do not map to the other domains]). Previous studies performed during the validation of the PBC-40 demonstrated strong correlations between individual fatigue domain and other symptom, cognitive, social and emotional domain scores (which, in turn, all correlated with each other), but no correlation with itch domain scores.11 To draw comparisons between patients with PBC and controls, all participants also completed the FIS, a generic fatigue measure that has been validated for self-completion in PBC and that, in contrast to the PBC-40, is suitable for use in normal control subjects.8
Sleep was assessed in all subjects, using both subjective and objective assessment protocols. Subjective sleep assessment was performed using a series of well-validated sleep assessment questionnaires. The Pittsburgh Sleep Quality Index (PSQI)12 assesses global sleep quality and disturbances in sleep patterns during the previous month. The PSQI sleep quality score has a potential score range of 0 to 21, with higher scores indicating worse overall sleep quality. The Epworth Sleepiness Scale (ESS, possible score range 0-24) assesses daytime hyper-somnolence, a score of 10 or more being indicative of significant daytime hypersomnolence.13 All patients were also asked to keep a sleep log in which they recorded the time they went to bed at night and got out of bed in the morning, the length of time they perceived that it took them to get to sleep, and the number of hours per night that they perceived that they had slept. Objective sleep assessment was performed by the use of accelerometry. All subjects wore an Actigraph activity monitor (MTI Actigraph; USA [model 7164: http://www.theactigraph.com]) for 7 days. Actigraphs are small, noninvasive, wrist-mounted devices containing a movement detector and downloadable memory. They have a pre-programmed activity detection mode that is able to discriminate between movement patterns associated with sleep and wakefulness.14 Previous studies have shown a close correlation between “sleep” and activity as discriminated by the Actigraph, and wake/sleep patterns determined by conventional polysomnography.14 Sleep/wake cycles (Supplementary Fig. 1; Supplementary material for this article can be found on the HEPATOLOGY website (http://interscience.wiley.com/jpages/0270-9139/suppmat/index.html) were scored in the current study using specific computer software obtained from the manufacturer (Actisoft) in a blinded manner by an observer unaware of the status of the subject (patient with PBC or control) or the corresponding subjective sleep assessment and fatigue severity data. Data were collected for the whole of each 24-hour assessment period. Individual data sets were then extracted for nighttime (24:00-06:00; a time period across which all subjects reported themselves to be in bed in their sleep logs), daytime (10:00-20:00; a time period across which all patients reported themselves as being out of bed (i.e., had risen for the morning, although clearly some went back to bed during this time for their daytime sleep) and morning/evening transition (06:00-10:00 and 20:00-24:00).
Domiciliary Sleep Studies.
Ten patients with more severe daytime sleepiness (ESS >14) were invited to undergo a domiciliary sleep study to exclude the obstructive sleep apnea syndrome (OSA) as a possible contributor to the sleepiness. The monitoring equipment (Embletta, ResMed, UK) comprised sensors recording oxygen saturation and heart rate (both by finger oximeter), airflow (nasal cannulae), thoracic and abdominal movements (inductance bands), snoring (neck vibration sensor), and body position. Patients were instructed in use of the equipment. They returned the equipment the following day for download and analysis. Sleep-breathing events were identified manually using conventional classification criteria and expressed as an apnea-hypopnea index (AHI) (i.e., the average number of apneas and hypopneas per hour of overnight time in bed. An AHI >15 events per hour was regarded as suggestive of OSA in this somnolent population.
Data were analyzed using Graphpad software (Prism). Comparisons were made between groups using Student t test (for normally distributed data sets) and Mann-Whitney non-parametric testing for non-normally distributed data sets. Correlations were by Spearman rank test. Proportions were compared using Fisher's exact test.
Subjective Sleep Assessment.
Patients with PBC spent a median of 30 minutes longer in bed at night than controls (Supplementary Table 1) but exhibited significantly greater sleep latency (30 minutes [range, 2-180] vs. 17.5 [5-120], P < .0005). Patients with PBC also perceived that they slept for a significantly shorter time per night than controls (6.1 hours [range, 3.5-10] vs. 7.0 [2.5-9], P < .05). Global sleep quality as assessed by the PSQI was significantly lower in the patient group than in the normal controls (Table 2). This reduction in sleep quality in patients with PBC was associated with a significant increase in daytime somnolence as assessed using the ESS (Fig. 1A, Table 2). More than 50% of patients with PBC had an ESS score of greater than 10 (an established threshold for significant daytime somnolence) compared with less than 15% of normal controls (Fig. 1B).
Table 2. Demographic Data and Results of Sleep and Other Symptom Assessment Tools From Subjects With PBC and Normal Controls
NOTE. Statistically significant results are shown in bold.
No significant difference was seen in mean total number of minutes of sleep per 24-hour period (as assessed by Actigraphy) between the patient with PBC and normal control groups (Table 3). When amount of time spent with a sleep-associated activity pattern was compared between the two groups for the daytime window, significantly higher sleep-pattern activity was seen for the patients than for the controls, with subjects with PBC sleeping, on average, for almost twice as long during the day (Table 3). This apparent increase in daytime sleep in patients appeared to be compensated for by a significant reduction in nighttime sleep (Table 3). Patients with PBC exhibited, therefore, a balanced shift away from nighttime sleep and toward daytime sleep. Reduced nighttime sleep in patients with PBC did not appear to result from an increase in the number of wakening episodes (Table 3).
Table 3. Objective Assessment of Sleep Performed by Actigraphy
NOTE. Nighttime sleep refers to numbers of minutes of sleep-pattern activity between 24:00 and 06:00. Daytime sleep refers to numbers of minutes of sleep-pattern activity between 10:00 and 20:00. Morning/evening sleep refers to numbers of minutes of sleep-like activity between 06:00 and 10:00 and between 20:00 and 24:00. All assessments were performed over 7 days, and the data refer to the relevant 7-day mean. Parameters with significant inter-group differences are denoted in bold.
Relationship Between Sleep Patterns and Fatigue Severity.
Having demonstrated significant abnormality in sleep patterns in patients compared to controls, we went on to study the association between sleep abnormalities and fatigue. A highly significant association was seen between the degree of daytime somnolence as assessed by the ESS and fatigue severity as assessed by the fatigue domain of the PBC-40 (Fig. 2A). A similar significant correlation was also seen between ESS and fatigue severity as assessed using the less sensitive FIS measure (Fig. 2B; confirmation of the association between fatigue and daytime somnolence using this second measure was important given that the PBC-40 fatigue domain includes a question specifically relating to daytime somnolence (the highlighting of this symptom during the qualitative phase of the PBC-40 derivation being one of the reasons for performing the current study), raising the possibility that the association between PBC-40 fatigue domain and ESS scores could be biased by shared content). In contrast to the strong association between daytime somnolence and fatigue in patients with PBC, no significant association was seen between overall sleep quality as assessed by the PSQI and fatigue severity (Fig. 2C). In keeping with the observation that fatigue severity showed an association with ESS scores, fatigue severity in patients was also found to show a strong association with actual amount of daytime sleep (Fig. 3). Although the patients had marginally higher BMI values than controls (29.0 ± 5.1 kg/m2 vs. 26.8 ± 4.9, P < .05), no association was seen between daytime somnolence as assessed using either subjective or objective measures and BMI in the PBC patient cohort (ESS vs. BMI r2 = 0.03, daytime sleep minutes vs. BMI r2 = 0.05, both P = NS). Amount of daytime sleep in patients with PBC showed no association with any parameter of liver disease severity, suggesting that increased daytime sleep activity was not a simple consequence of the complications of advancing liver disease (Table 4).
Table 4. Correlations Between Biochemical and Hematological Parameters of Liver Function and Daytime Sleep Minutes as Assessed by Actigraphy
Statistically significant results are shown in bold.
Comparison between the control group and subgroups of patients with the highest and lowest fatigue scores (the quartiles of patients with the highest and lowest PBC-40 fatigue domain scores) suggested that low-fatigue patients with PBC had both objective and subjective parameters of daytime sleep that were identical to controls, with abnormality in daytime sleep patterns being entirely restricted to the high-fatigue group (Fig. 4).
The associations between objective and subjective measures of sleep and the other symptoms of PBC assessed using the domains of the PBC-40 are outlined in Table 5A. The additional strong associations between the ESS and the cognitive, other symptoms, and social and emotional domains of the PBC-40 are likely to reflect the strong overlap seen between these domains and the fatigue domain.11 No association was seen between any parameter of sleep assessed objectively by Actigraphy and itch severity (Table 5B).
Table 5A. Correlations Between Individual Symptom Domains of the PBC-40 and Sleep Determined by Questionnaire (Pittsburgh Sleep Quality Index [PSQI] and Epworth Sleepiness Scale [ESS] and Inactivity From Actigraphy
PBC-40 Domain Name
Mean Daytime Sleep
NOTE. The details of the individual PBC-40 domains and their inter-relationships are described in detail elsewhere.1, 1 Values presented are r2,
Table 5B. Correlation Between Itch Severity as Assessed Using the Itch Domain of the PBC-40 and Sleep Parameters Assessed Using Both Objective Assessment Protocols Table 5A.
NOTE. The details of the individual PBC-40 domains and their inter-relationships are described in detail elsewhere.1, 11 Values presented are r2,
Total sleep (min/24 h)
Nighttime sleep (min)
Daytime sleep (min)
Number of wakenings per hour at night
Domiciliary Sleep Studies in PBC.
Given the previously recognized association between daytime somnolence, fatigue, and OSA, we set out to examine the extent to which OSA contributed to sleep abnormality and fatigue pathogenesis in a small subgroup of patients with PBC with ESS scores higher than 14. 9 of the 10 patients invited to take part participated in this phase of the study (Table 6). They were of similar age (mean, 57.1 ± 7.2) and BMI (27.0 ± 3.7) to the PBC patient cohort as a whole (P = NS vs. whole PBC cohort for both parameters). Unsurprisingly, given the association between ESS and fatigue severity, PBC-40 Fatigue domain scores were significantly higher in the patients participating in the domiciliary sleep studies than in the PBC patient cohort as a whole (42.4 ± 6.3 vs. 32.5 ± 1.4, P < .01). Of the nine patients studied, one had an AHI of 41 events/hour, suggesting coexisting OSAS, whereas the other eight all had AHI <10 (Table 6).
Table 6. Findings From Nine Patients With PBC and Epworth Sleepiness Scale (ESS) Values Greater Than 14 Who Participated in Domiciliary Sleep Studies
PBC-40 Fatigue Domain Score
NOTE. A/H index denotes number of apnea/hypopnea episodes per hour.
In this study we set out to examine patterns of sleep in patients with PBC and to characterize the associations between abnormalities in sleep pattern and severity of fatigue. This study was prompted by three earlier observations: First, associations between global sleep abnormality and fatigue severity have previously been suggested (but not confirmed) in PBC.3 Second, clinical descriptions suggestive of sleep abnormality featured prominently during the qualitative phase of the PBC-40 derivation.1 Third, abnormalities of circadian rhythm, including, in particular, increased daytime somnolence, have recently been described in other fatigue-associated conditions and implicated heavily in the pathogenesis of fatigue.9, 10 The data presented here confirm the presence of significant sleep abnormality in PBC, highlight a particular increase in daytime somnolence, and suggest that the degree of such daytime somnolence correlates significantly with severity of perceived fatigue.
In demonstrating a sleep abnormality (quantified in the current study using complimentary subjective and objective measurement approaches), the degree of which correlates significantly with severity of fatigue, our study confirms, and significantly builds on, the findings of the only other study in the literature addressing the question of sleep in PBC.3 That Canadian study also showed an apparent association between sleep disturbance and fatigue in PBC. It was, however, limited in terms of the scope of the sleep assessment protocol adopted and the approach taken to fatigue assessment. In the current study, we extended the previous work by using a questionnaire specific for daytime sleep symptoms (the Epworth Sleepiness Scale [ESS]), as well as an objective assessment of inactivity that has been shown to correlate well with sleep using polysomnography.14 Both correlated with fatigue severity as assessed using, for the first time in this context, a QOL measure specifically designed and optimized for use in PBC.1, 11
In patients with PBC, there was no difference in overall amount of sleep taken over the full 24-hour cycle. However, by considering predetermined time windows, we were able to examine changes in nighttime and daytime sleep patterns. Using this approach, we found that, in comparison with age- and sex-matched controls, patients with PBC exhibited a significant reduction in actual nighttime sleep (although they reported that they spent a similar length of time in bed to controls) and a significant increase in daytime sleep; findings that are consistent with the abnormally high ESS scores seen in the group with PBC. The reduction in nighttime sleep seen in patients with PBC did not occur as a result of an increase in numbers of wakening episodes. The obvious implication, that the difference between patients with PBC and controls was with regard to the length of time spent awake after each wakening episode, together with the self-reported increase in length of time taken to get to sleep in the first place in patients with PBC, would point to difficulty in initiating sleep at night as the explanation for reduced nighttime sleep time in patients with PBC. The presence of itch in the patients with PBC appeared to play no role in reducing nighttime sleep.
The questions of what is the cause of the sleep abnormality in PBC and how, in mechanistic terms, the phenomena of fatigue and sleep abnormality are linked are important ones, the answers to which may influence the future clinical management of patients with PBC. One possibility is a simple “cause and effect” model in which one of the phenomena is a direct consequence of the other (i.e., the process is driven by abnormal nighttime sleep patterns, with daytime somnolence and fatigue being simple consequences). The lack of association between PSQI parameters of sleep other than those relating specifically to daytime sleep and fatigue would, however, argue against daytime somnolence and fatigue being simple consequences of poor nighttime sleep. A second possibility is that the patient group has a greater prevalence of another pathological process that is itself associated with daytime somnolence and fatigue. The most obvious potential confounding condition would be the obstructive sleep apnea syndrome, a condition well recognized to be associated with both daytime somnolence and fatigue. The possibility of OSA is further suggested by the relatively high BMI values seen in the populations participating in this study. However, the domiciliary sleep studies performed in nine of the patients with PBC with significant daytime somnolence (ESS values > 14) showed only one with significant OSA, which argues strongly against OSA being the cause of daytime somnolence and fatigue in the majority of fatigued patients with (although clearly sporadic cases of OSA are likely to arise in the population with PBC just as in the population without PBC). The third, and we would argue most plausible, possibility is that a common central pathogenetic process gives rise to inter-related symptoms of fatigue and sleep abnormality.
This model is fully consistent with previous observations made in both animal models of cholestasis (notably the bile duct–resected rat) and in human patients with PBC. Studies performed in animal models of cholestasis have suggested links between central nervous system changes resulting from cholestasis and fatigue-associated behavior patterns,15, 16 whereas magnetic resonance (MR) studies performed in human PBC have shown cholestasis-associated brain stem changes, the degree of which correlates strongly with fatigue severity.17 A potential mechanism for the cholestasis-associated effects on both fatigue and sleep is suggested by the observation that, in cholestatic animal models, changes in brain inflammatory cytokine levels (possibly resulting from the recently demonstrated increase in central nervous system infiltrating cytokine releasing mononuclear seen in such animals18) are strongly linked to fatigue-like behavior. Interleukin-6 is an inflammatory cytokine present at elevated levels in human liver with PBC19, 20 and is excessively released by both cultured biliary epithelial cells in response to hydrophobic bile acids of the type retained in the liver in cholestasis21 and spontaneously by PBMC from patients with PBC.22 Interleukin-6 plays a key role in controlling circadian sleep patterns and promoting fatigue23 and is elevated (together with tumor necrosis factor alpha and interleukin-1) in a number of conditions associated with daytime somnolence and fatigue of non-liver origin.24
In addition to raising important questions regarding the pathogenesis of fatigue in PBC, the observation of a strong link between fatigue and daytime somnolence gives rise to a key clinical question: does therapy effective at reversing daytime somnolence also result in re-duction in fatigue (a key therapeutic aim in the clinical management of PBC)? The recent observation that, in a small case series, the drug modafinil, an agent used to treat daytime somnolence in conditions such as narcolepsy, is effective in reducing fatigue in PBC would suggest that this may be the case.25 Our observation of a high incidence of daytime somnolence, the degree of which strongly associates with severity and fatigue, together with this pilot observation that therapy effective at reducing daytime somnolence may be effective in reducing fatigue, strongly suggest that further, more formal therapeutic trials of agents such as modafinil aimed at ameliorating fatigue in PBC via effects on daytime somnolence are warranted.