Actigraphy assessments of circadian sleep-wake cycles in the Vegetative and Minimally Conscious States

The Vegetative State (VS) or Unresponsive Wakefulness Syndrome (UWS [1]) is thought to reflect the dissociation of the two primary components of consciousness - awareness and wakefulness [2, 3]. A common tool for the assessment of awareness is the Coma Recovery Scale Revised [4] (CRS-R) which includes subscales designed to assess a range of functions, including auditory, visual, motor, verbal, communication and arousal. A brain-injured patient is considered to possess awareness if they produce non-reflexive responses to stimulation, such as tracking an object that is moving in front of the eyes, or following a verbal command. Patients in the VS do not produce non-reflexive behavior and are, therefore, considered to lack awareness [5, 6]. Patients in the Minimally Conscious State (MCS) exhibit some reproducible but inconsistent signs of awareness, although communication remains absent [6, 7].

Wakefulness, on the other hand, is thought to be preserved in both VS and MCS patients. According to the standards for VS and MCS outlined by the Multi-Society Task Force for Permanent Vegetative State [5] and the Royal College of Physicians [6], 'wakefulness' refers to the presence of typically cycling periods of eye-closure and eye-opening that give the appearance of sleep-wake cycles. While a great deal of behavioral and neuroimaging research has focused on the assumption of unawareness in these patients [4, 8–10], very little is known regarding the assumption of preserved sleep-wake rhythms.

A typical sleep-wake cycle follows a circadian rhythm, with a period of between 19- and 28-hours [11]. Electroencephalography (EEG), in combination with other physiological measures as part of polysomnography, is the gold-standard approach for the assessment of sleep-wake cycles [12]. However, the results of the limited EEG investigations of circadian sleep-wake cycling in VS and MCS patients are inconsistent with the assumption of preserved wakefulness. Landsness et al. [13] observed sleep-wake-like changes in the EEG of six MCS patients across one day, while the EEG of five VS patients remained unchanged between periods of eye-opening and eye-closure. Isono et al. [14] also reported an absence of EEG sleep-wake changes in 4 out of 12 VS patients. High variability has also been observed in other physiological circadian rhythms in VS and MCS, including body temperature and hormone levels [15, 16], blood pressure and heart rate [16, 17], and sleep-related erections [18]. Circadian-like variations in arousal have also been reported in both VS and MCS patients, as indexed by fluctuating behavioral abilities across the day [19]. Bekinschtein et al. [20] observed well-formed circadian rhythms in the body temperatures of two VS patients with traumatic brain injuries (TBI), but absent rhythms in three VS patients who had sustained non-traumatic brain injuries (non-TBI), indicating the potential relationship between etiology and circadian rhythms. It appears, therefore, that, contrary to the diagnostic guidelines describing these conditions, a great deal of variability exists both within and across VS and MCS patient groups with regard to the relative preservation of circadian rhythms.

An indirect and inexpensive approach to detecting circadian sleep-wake cycles from large numbers of patients is wrist actigraphy, in which a wrist-mounted device is used to record the frequency and amplitude of motor activity [12]. This method is known to correlate well with polysomnographic measurements of sleep and wakefulness in healthy individuals, as well as non-ambulatory patients, such as those with C5 to C7 tetraplegia [21–23]. A number of algorithms have been developed in order to produce minute-to-minute estimations of sleeping/waking from short-term variations in actigraphy data in healthy individuals. Broadly, these algorithms judge an individual to be awake or asleep at a given sample point by weighting the amount of movement in a number of preceding sample points by a set of predefined constants. Such approaches have reported between 88 and 97% concordance with polysomnography in healthy individuals (see [21] for a full review). However, none of these approaches have been validated with VS or MCS patients by means of concurrent polysomnography and actigraphy recordings. Nevertheless, a circadian sleep-wake rhythm - that is, more activity during waking hours and less activity during sleeping hours - can be readily identified from raw actigraphy recordings, and makes fewer assumptions than these un-validated algorithms (for example, [24, 25]). In the only article to report actigraphy-based assessments of sleep-wake rhythms in VS, Bekinschtein et al. [26] described a greater deterioration in the circadian rhythmicity evident in the actigraphy of one VS patient relative to an MCS patient. De Weer et al. [27] also reported day-night variation in the amount of movement (as measured by actigraphy) in two TBI MCS patients, but not in a non-TBI MCS patient. However, in neither of these studies was circadian rhythmicity examined statistically.

In order to investigate the relative preservation of circadian sleep-wake rhythmicity in patients in the VS and MCS, we recorded wrist actigraphy from 55 patients (18 VS, 37 MCS) across four days, and subjected the data to cosinor rhythmometry analyses (see Methods), a standard statistical approach for circadian rhythm identification. By definition, all of these patients are considered to possess circadian sleep-wake cycles [5–7]. In keeping with the studies described above, however, we expected to see variability in the extent to which circadian sleep-wake rhythms were preserved across patients as a function of etiology (TBI vs. non-TBI) and diagnosis (VS vs. MCS). We also predicted significant relationships between the behavioral profiles of these patients - as indexed by their CRS-R subscales - and the relative preservation of their circadian sleep-wake rhythms.

A total of 46 out of the whole group of 55 patients (84%) exhibited significant 24-hour rhythms in their actigraphy data after permutation testing. This proportion is significantly lower than the diagnostic expectation that all patients retain significant circadian rhythms (Fisher's Exact Test, P <.01). When separated according to diagnosis, 15/18 VS patients (83%) and 31/37 MCS patients (84%) returned circadian rhythms that passed this statistical test. When separated according to etiology, 24/31 TBI patients (77%) and 22/24 non-TBI patients (92%) exhibited circadian rhythms. There was no significant effect of diagnosis or etiology on the proportions of patients exhibiting circadian rhythms (Fisher's Exact Tests, all P >.14). While age significantly differed across etiologies, it did not significantly correlate with any of the four rhythmicity variables (mesor, amplitude, acrophase or goodness-of-fit, as indexed by the log-transformed zero-amplitude F-ratio).

VS versus MCS patients

Four one-way ANOVAs with diagnosis (VS, MCS) as the factor of interest revealed the main effects of mesor (F(1,54) = 4.441, P <.05), amplitude (F(1,54) = 6.819, P <.05), and goodness-of-fit (F(1,54) = 16.517, P <.001), but not acrophase. Together these reflect the greater average amount of movement across the four days (mesor), the greater amount of movement during periods of activity (amplitude), and greater statistical reliability of the circadian rhythms (goodness-of-fit) of MCS patients relative to VS patients (see Figure 1).

Due to the high inter-correlations between these three significant rhythmicity variables (all absolute r >.33), all three variables were entered into a backward stepwise logistic regression in order to determine their relationships with diagnosis, over and above the effects of the other two variables. This regression retained only goodness-of-fit in the model as a significant predictor (Wald = 10.189, Beta (SE) = -2.043 (.640), P <.01), indicating significantly weaker circadian rhythms in VS patients relative to MCS patients, regardless of the amount of movement produced by these patients.

On the basis of periodic eye-opening and eye-closure, patients in the VS and MCS are considered to have preserved circadian sleep-wake rhythms [5, 6]. However, by means of an indirect measure of sleep-wake rhythmicity - wrist actigraphy - we have shown that a significant proportion of these patients do not exhibit statistically reliable circadian sleep-wake rhythms. The observed variability across patients is consistent with previous smaller studies of circadian rhythmicity in VS and MCS (see Background), and is the first evidence from a large-scale study of sleep-wake cycling using the inexpensive and non-invasive method of wrist actigraphy.

While there was no significant difference in the proportion of patients exhibiting significant sleep-wake rhythms between VS and MCS patients, the goodness-of-fit of the circadian rhythms in the data of MCS patients were significantly higher than those of the VS patients (see Figure 1). This result indicates that the circadian sleep-wake cycles of MCS patients were significantly more statistically reliable than those of VS patients. Importantly, this remained true when taking into account the morphology of the rhythm (that is, its mesor and amplitude), indicating that the effect of diagnosis on the statistical reliability of the circadian rhythms is not driven by simple differences in the amount that a patient moves, but rather reflects differences in the circadian rhythmicity with which this movement occurs.

A master biological clock in the hypothalamic suprachiasmatic nuclei (SCN) is considered to maintain the timing of circadian rhythms. The SCN in turn modulates the activity of the ascending reticular activating system (ARAS) - a circuit of subcortical nuclei responsible for promoting wakefulness (see [29] for a review). One region of the ARAS - the central thalamus - is known to be crucial for the regulation of arousal and has been linked to the disorders of consciousness exhibited by VS and MCS patients [30]. Indeed, the extent of atrophy in this region of the thalamus has been associated with the degree of disability exhibited by these patients [31]. More broadly, greater thalamic atrophy has been observed in VS patients relative to MCS patients using in vivo diffusion tensor imaging [32] (DTI). The weaker circadian sleep-wake rhythms observed in the VS patients in the current study are, therefore, entirely consistent with these differential patterns of damage to the thalamus.

Etiology was also shown to have a small effect on the amount that patients moved during periods of activity (amplitude) and the statistical reliability of the circadian rhythm (goodness-of-fit). Similarly, Bekinschtein et al. [20] observed reliable circadian temperature rhythms in only TBI VS patients, but not in non-TBI patients, while De Weer et al. [27] detected sleep-wake activity changes in only TBI MCS patients. The primary neuropathology associated with TBI is diffuse axonal injury with relative preservation of the cortex, while non-TBI involves more widespread damage to the cortex and basal ganglia [33–39]. The greater impairment of circadian rhythms in non-TBI patients relative to TBI patients reported here is, therefore, consistent with the general patterns of neuropathology associated with the two etiologies. Indeed, mouse models of hypoxic brain injury have been shown to result in impaired sleep-wake cycling [40].

Significant relationships were also observed between the behavioral profiles of the patients - as indexed by their CRS-R sub-scales - and aspects of their circadian rhythmicities. A significant positive relationship was found between the motor subscale and the mesor and amplitude of the rhythm. The motor subscale of the CRS-R is scored from flaccid motor tone at its lowest, to object manipulation and automatic motor responses at its highest (before emergence from MCS). Since wrist movements were used to indirectly measure the circadian rhythms, it is unsurprising that greater amounts of movement exhibited by patients across the recording period (mesor, amplitude) are related to their overall abilities to produce motor output during behavioral assessments. This result suggests the need for caution in the use of actigraphy for assessing circadian sleep-wake rhythms since they rely on motor output for a rhythm to be detected. Nevertheless, our analyses have demonstrated that significant changes in the statistical reliability of the rhythms across diagnoses are not dependent on the amount of movement produced, suggesting that actigraphy can be used to assess the statistical reliability of circadian sleep-wake cycles, regardless of the degree of activity exhibited by the patients.

A combination of the visual and motor subscales significantly predicted the goodness-of-fit of the circadian rhythms. The visual subscale score describes behaviors from absent visual startle at its lowest, through fixation and pursuit, to object recognition at its highest. This relationship is of particular interest since the master clock for circadian rhythms, the SCN, is itself timed by light inputs from the retina during the day, as well as melatonin from the pineal gland at night [29]. The more purposeful eye-movements of those scoring high on the visual subscale may allow for differing levels of light to reach the retina - perhaps through a greater ability to orient toward light or to maintain eye-opening for longer periods - and, consequently, result in a strengthening of the rhythm via the SCN. The predictive value of the visual subscale could, therefore, be considered to be consistent with our understanding of the biological generators of sleep-wake rhythmicity. This conclusion is necessarily speculative, however, since it is unclear whether high visual functioning is associated with a greater degree of orientation toward light or longer periods of eye-opening. Further investigation of this relationship will contribute to our understanding of the exogenous cues that drive circadian rhythms in VS/MCS patients.

A significant relationship was also found between the visual subscale of the CRS-R and the acrophase (time of maximal activity) of the rhythm, over and above the contribution of the other CRS-R subscales, or of the diagnosis of the patient. The relationship with acrophase reflects the tendency for patients with higher visual functioning to be most active later in the afternoon than patients with lower visual functioning (Visual Score >= 1, Mean acrophase (SD) 18:20 (three hours); Visual Score = 0, Mean acrophase (SD) 17:20 (two hours)). Consistent with this observation, exposure to higher levels of light has been associated with later peaks of activity in institutionalized individuals [41, 42]. However, the activity peaks of healthy individuals occur earlier in the day than those observed in the patients here, typically between approximately 13:30p.m. and approximately 16:00p.m. [24]. It has been observed that the levels of light experienced by institutionalized patients are considerably lower than those of non-institutionalized individuals [42, 43], and since the patients in the current study were residing on a hospital ward during the recording period, it is likely they were exposed to abnormally fluctuating levels of light compared with healthy individuals. Unfortunately, we were unable to record light levels alongside actigraphy; however, future studies investigating their contribution to the timing of activity of VS and MCS patients will be invaluable.

Since we inferred the circadian rhythms of patients from wrist actigraphy, it is likely that the recordings contain some levels of exogenous activity, perhaps from nurses moving the patient from bed to chair. Since these patients were all admitted to the same ward of the University Hospital of Liège as part of the same research protocol, they all received equivalent levels of care and were involved in the same assessments throughout the day - for example, behavioral tests, PET and MRI. As a result, the potential exogenous noise in the data would then be equally distributed across all patients. Our conclusions regarding the effects of diagnosis, etiology and behavioral profile on sleep-wake cycles, therefore, would remain valid despite this potential confound. The use of simultaneous video-recordings would allow for the exclusion of activity that is generated exogenously and would further validate our findings.

A final caveat is that the apparent absence of reliable circadian rhythms in some of our patients may be a result of a lack of sensitivity of the actigraphy method, rather than the true absence of those rhythms. While wrist actigraphy has been validated for sleep assessment in patients with C5 to C7 tetraplegia [23], these patients are nevertheless capable of small but purposeful wrist movements. Patients in the VS, however, are by definition unable to produce purposeful movements, although spontaneous movements are common. Similarly, due to the heterogeneity of brain injuries of these patients, it is not clear whether the presence of actigraphy-detected waking is necessarily indicative of concurrent cerebral waking. Future validation of the relationship between polysomnography and actigraphy measures of sleeping and waking in VS and MCS patients is needed in order to fully characterize the nature of their circadian rhythms.

Our analyses indicate a greater impairment of circadian sleep-wake cycling in patients in the VS compared with those in the MCS, and in those with non-TBI compared with TBI. The significant differences observed between VS and MCS patients support the conclusion that these are diagnostically distinct entities. However, they also suggest that despite periods of eye-closure and eye-opening, sleep-wake cycles are not necessarily present despite the clinical criteria for these conditions [5–7]. Wrist actigraphy is considerably less expensive and less invasive than other forms of sleep-wake monitoring and may, therefore, provide a reliable means of determining the extent to which these cycles are preserved in individual patients. These recordings could also allow clinicians and researchers to identify the time of day in which a patient is most active, in order to schedule behavioral and/or neuroimaging assessments for a time that maximizes the likelihood of detecting an appropriate response (see [26]). Future validation of the relationship between actigraphy and polysomnography measures of sleeping/waking in VS and MCS patients will allow for a more complete understanding of the physiological nature of these circadian rhythms. Follow-up studies will also determine the prognostic utility of wrist actigraphy for VS and MCS patients.