Actigraphy is based on the principle that sleep is characterized by little or no movement, whereas wake is marked by activity (Ancoli-Israel et al., 2003; Sadeh, 2011). Actigraphs contain an accelerometer which measures movement continuously during 24-hours, for several days or weeks (Ancoli-Israel et al., 2003; Sadeh, 2011). Inactivity moments recorded by the actigraph can be helpful to infer sleep and thus study sleep-wake rhythms (Ancoli-Israel et al.
, 2003). Once the actigraph data is downloaded, wake and sleep periods are automatically scored by computer algorithms and sleep variables estimated: total sleep time (TST), total wake time, sleep efficiency (SE), number of awakenings during sleep, fragmentation index and sleep onset latency (SOL) (Ancoli-Israel et al., 2003). More recent devices look like wristwatches, are water resistant and have extra features as light and skin temperature sensors (Ancoli-Israel et al., 2003; Sadeh, 2011). Actigraphy has been widely used in sleep medicine and research, especially for being a non-invasive, cost-efficient and well-tolerated assessment tool. This is reflected in the high rate of growth in the number of publications over the last two decades (Sadeh, 2011). In 2007, the American Academy of Sleep Medicine (AASM) developed evidence-based recommendations for the use of actigraphy, indicating it as an assessment tool in the evaluation of specific sleep disorders and their response to treatment: sleep apnoea, insomnia, hypersomnia and circadian rhythm disorders (CRD) (Morgenthaler et al.
, 2007). Furthermore, it recognises actigraphy as a reliable and valid method to estimate TST in appropriate populations, where PSG is not available or cannot be performed (Morgenthaler et al., 2007).The validity and reliability of actigraphy are usually debated based on comparisons against more or less objective methods, as PSG and sleep diaries respectively (Ancoli-Israel et al., 2003; Sadeh, 2011). There is nowadays a wide variety of commercial devices available in the market which, similarly to actigraphy, measure activity levels and infer sleep (Montgomery-Downs et al., 2012; Mantua et al.
, 2016; Lee et al., 2017). Surely, there are advantages and disadvantages associated with each of them. This essay is then going to discuss the diagnostic and practical value of actigraphy against these three methods. Actigraphy versus Commercial DevicesBoth actigraphy and commercial devices monitor activity continuously and infer sleep.
There is an increasing adherence to commercial devices by the population and, due to this and their intrinsic advantages, these commercial devices are currently being tested as an alternative to actigraphy (Montgomery-Downs et al., 2012; Mantua et al., 2016; Lee et al.
, 2017).An early study by Montgomery-Downs et al, 2012 reported that a Fitbit device and actigraphy differed significantly from each other and from PSG. Both methods presented similar TST and overestimated SE, being more evident for the Fitbit device, which overestimated SE by 5.2% compared to actigraphy (Montgomery-Downs et al., 2012). Even though both presented high sensitivity (ability to accurately infer sleep) and poor specificity (ability to infer wake), the Fitbit showed slightly higher sensitivity in all sleep stages and during arousals (Montgomery-Downs et al., 2012). A strong point of this study was comparing the data against PSG; However, the short duration of the study (one night) did not take into account the ‘first-night effect’ on PSG (Montgomery-Downs et al.
, 2012; Lee et al., 2017). Also, it was not stated which Fitbit device had been used. Conversely, a very recent study by Lee et al., 2017 showed that sleep start times and circadian rest-activity cycles recorded over 14 days by Fitbit Charge HR and Actiwatch2 did not differ and were significantly correlated, in a healthy adult population. However, the sleep duration measured by the Fitbit charge HR tended to be consistently overestimated each day by 20-30 minutes compared to Actiwatch2 (Lee et al.
, 2017). One strength of this study is that Fitbit data was not diary-adjusted since the device is set to detect sleep automatically (Lee et al., 2017). Also, the authors went on comparing activity measurements (number of steps and activity score) and these turned out similar (Lee et al.
, 2017). In a broader picture, Mantua et al., 2016 attempted to validate many commercial devices by comparing them against PSG and Actiwatch Spectrum. TST did not differ significantly from the actiwatch, but SE was different among all devices (Mantua et al., 2016). Only the SE determined by actiwatch was correlated to PSG. However, it is important to bear in mind that different commercial devices may use additional or different physiological signals to infer sleep, apart from movement (De Zambotti et al., 2016).
This may then explain some inconsistency in the results across studies using different commercial devices (Lee et al., 2017). For instance, the Fitbit charge HR infers sleep also based on HR measurements and this can affect the reliability and accuracy of data in participants with cardiac arrhythmias (De Zambotti et al., 2016). Another drawback of commercial devices is that some do not seem to recognise sleep automatically, requiring the user to initiate that mode manually, making the data more subjective and vulnerable under unsupervised conditions (Mantua et al., 2016). Conclusions: Commercial devices may be of interest in more qualitative studies, where clinical accuracy is not essential (e.g.
therapeutic interventions), since the above results suggest similar TST among them, actigraphy and PSG (Mantua et al., 2016) and a possible ‘correction factor’ (Lee et al., 2017). When compared to actigraphy, commercial devices are of lower cost and can provide a wider range of physiological data in real-time, which may be beneficial in monitoring treatment efficacy (Liang and Ploderer, 2016; Lee et al., 2017). Additionally, it may encourage patients to be more actively involved in their therapy and ultimately be their own clinicians (Liang and Ploderer, 2016). From a research point of view, commercial devices may facilitate the development of global epidemiologic studies. Even though this market is evolving very quickly, more validation must be provided (Mantua et al.
, 2016).Actiwatches versus Sleep diariesSleep diaries are a direct way of getting information about one’s sleep hygiene and pattern (Morgenthaler et al., 2007). Moreover, it can be used to make sense out of actigraphy data when completed simultaneously (Morgenthaler et al.
, 2007). Currently, there are several studies showing a good correlation between actigraphy and sleep diaries (reviewed in Sadeh, 2011), but some exceptions need to be considered. For instance, a study by Marino et al., 2013 demonstrated that actigraphy is less accurate in classifying wake in insomniac patients. This may be due to the fact that insomniac patients tend to lie in bed quietly for long periods, but still awake – actigraphy will tend to overestimate TST and underestimate SOL (Meagher et al., 2013). In this particular case, sleep diaries can be helpful to adjust actigraphy data so it can be interpreted accordingly (Morgenthaler et al.
, 2007).In spite of the above mentioned, there is a likelihood for one to misperceive their sleep when completing the sleep log (Van Den Berg et al., 2008).
Indeed, Van Den Berg et al., 2008 compared the data obtained through sleep diary and actigraphy in a large sample (n=969, 6004 valid nights) and concluded that sleep diary estimates of TST were often higher than actigraphy. Interestingly, this difference was bigger in participants who had slept less than 5h recorded by actigraphy (Van Den Berg et al., 2008). Moreover, they showed that subjects with self-reported depressive symptoms reported a lower TST in the sleep diary when compared to non-depressed subjects (Van Den Berg et al., 2008).
Higher age and self-reported poor sleep quality and cognitive functioning were also associated with higher sleep diary TST estimates (Van Den Berg et al., 2008). However, it would have been interesting to compare the obtained data to a more objective method (PSG). Also, it is important to bear in mind that depression symptoms and cognitive impairment were self-reported and not assessed by a clinician (Van Den Berg et al., 2008). Nevertheless, these findings showed that actigraphy can be used as a tool to correct sleep misperception (Van Den Berg et al., 2008).
Conclusions: The aforementioned suggests that both methods may complement each other and therefore should be seen together to improve data accuracy (Morgenthaler et al., 2007). The completion of a sleep log whilst performing actigraphy is indeed recommended (Morgenthaler et al., 2007).Actiwatches versus PSGPSG is the gold-standard method to measure sleep objectively, in terms of its efficiency, duration, and onset (Morgenthaler et al., 2007; Montgomery-Downs et al., 2012).
However, the high-cost, procedure and time needed to perform and analyse the data are some of the main drawbacks associated with PSG (Marino et al., 2013). These promoted the searching for alternative tools. Reliability and validity of actigraphy compared to PSG were addressed by several studies over the past two decades (Ancoli-Israel et al.
, 2003; Sadeh, 2011). In a retrospective study, Marino et al., 2013 assessed actigraphy validity in detecting sleep and wakefulness against PSG in 77 participants, 22% with chronic primary insomnia.
The authors concluded that actigraphy performs well in classifying sleep (sensitivity>90% every participant), but it is less reliable in detecting wake (specificity=33% highly variable across participants) (Marino et al., 2013). Interestingly, actigraphy accuracy was lower in insomniac subjects with low sleep efficiency and even lower in older insomniac patients (Marino et al., 2013). This is related to when insomniac patients lie in bed motionless, but awake; In addition, there is an increase of wake proportion with aging, which may then explain the effect of age in the actigraph’s accuracy (Marino et al.
, 2013). Although one strength of this study is the variability within the sample (younger to older, sleep restricted and insomniac subjects), a couple of drawbacks can be noted. Firstly, it is the fact that data is from one night spent in a sleep laboratory and this did not take into account the “first-night effect” on PSG data (Montgomery-Downs et al., 2012; Lee et al., 2017). Secondly, PSG studies were scored by various technologists and thus are subjected to inter-scorer variability, especially in the determination of SOL (Rosenberg and Van Hout, 2013).A recent study by Mantua et al.
, 2016 in 40 healthy participants, showed that TST measured by actigraphy correlated well with PSG, but a weaker correlation was described for SE. Conclusions: These findings suggest that actigraphy can be a valid tool to estimate TST and useful when differentiation between sleep stages is not clinically relevant (Morgenthaler et al., 2007; Sadeh, 2011; Marino et al., 2013). Even though PSG is compulsory for the diagnosis of most sleep disorders, actigraphy can provide relevant data to support diagnosis (e.g. CRD and hypersomnia) (Morgenthaler et al., 2007).
In these contexts, actigraphy can be used to record one’ sleep-wake cycles over long periods, under their natural environment, in a low-cost and more simple manner (Morgenthaler et al., 2007).ConclusionAmong other variables, actigraphy measures activity and this can be used to evaluate sleep-wake rhythms. Over the past two decades, actigraphy was used and tested in the assessment of sleep disorders and evaluation of treatment responses (Sadeh, 2011). It is though difficult to find a general consensus about actigraphy, since each study tests different actigraph(s), using different algorithm(s) (Sadeh, 2011).
The same notion applies to the commercial devices (Mantua et al., 2016). Nevertheless, reliability and validity of actigraphy were tested against PSG and, overall, findings suggest that actigraphy can be useful, being more accurate in classifying sleep than wake epochs (Sadeh, 2011; Marino et al., 2013; Mantua et al.
, 2016). Also, it can complement other more subjective/objective measurements (Morgenthaler et al., 2007).
There are other methods to further assess one’ sleep complaints and help to establish a diagnosis. The selection of the most appropriate method depends on the clinician, patient, clinical interview and clinical question to be answered (Morgenthaler et al., 2007). Each of the addressed methods has advantages and limitations and these should be taken into consideration not only when selecting the most appropriate, but also when interpreting the data (Morgenthaler et al., 2007). In addition, it is important to bear in mind that there are different levels of objectivity across these methods and, therefore, they are not interchangeable (Morgenthaler et al., 2007).