Actigraphy versus Polysomnography to Identify Changes in the Sleep Patterns of Adults with Insomnia

Introduction. Insomnia is the most common sleep-wake disorder worldwide. Polysomnography is the complementary exam considered the gold standard for the identification of sleep disorders. However, Actigraphy – a method easily applicable for several consecutive days –, has emerged as an alternative tool. Objective. To evaluate the effectivity of actigraphy in comparison with polysomnography, in determining the sleep patterns of adults with insomnia. Method. Systematic review with meta-analysis, on which randomized observational and clinical studies were searched in the Cochrane Library, on MEDLINE through the PUBMED Portal, in the Google Scholar and on the TRIP DATABASE meta-search engine, from November 2019. The GRADE approach was used to evaluate the quality of the evidence. The meta-analysis was performed using the Inverse-variance weighting, considering the 95% confidence interval (95% CI). Results. Six studies were included, representing 399 patients. The studies examine the effectivity of actigraphy in identifying the total sleep time, the latency, the Wake After Sleep and the sleep efficiency. It was noticed that the Actigraphy was close to polysomnography only when verifying sleep latency, with a small average difference of -1.46 (95% CI: -9.61 to 6.70 min.). The other outcomes evaluated had their values underestimated by actigraphy. The quality of the evidence was moderate and low. Conclusion. The evidence to evaluate the effectivity of actigraphy is still limited – especially when it is intended to compare it with polysomnography, since the studies found shown methodological limitations, mainly in the measurement of results.


INTRODUTION
Insomnia is the most frequent sleep-wake disorder in the world. In industrialized societies, its prevalence varies from 10% to 15% in the adult population and approximately 50% of them have a chronic course of the disorder 1 .
Insomnia is characterized by the difficulty in initiating or maintaining sleep, which has diurnal consequences not attributable to environmental circumstances or an inadequate way of sleeping 2 .
According to the Fifth Edition of the Manual of Mental Disorders 3 and Sateia et al 2 , chronic insomnia is defined by the difficulty related to sleep that appears at least three times often a week for three months consecutively.
Epidemiological studies have shown that insomnia is an independent risk factor associated with an increase in the number of accidents at work, general low impairment at work, presenteeism and absenteeism 4,5 .
Another study also found an increase in the prevalence of hypertension associated with short sleep (less than 6 hours) in individuals with insomnia 6 . In addition, studies have found a strong association of insomnia with lower quality of life related to physical and mental health 4,7 .
The diagnosis of sleep disorders, specifically insomnia, is obtained through a good anamnesis, a subjective assessment based on questionnaires and sleep diaries and an objective assessment by the Multiple Sleep Latency Test (MSLT), the Alarm Clock Test, the Polysomnography (PSG) and the Actigraphy (ACT) 8 . Studies that compared ACT's performance with PSG's performance to assess its precision and specificity identified high precision and sensitivity. The difficulty was in terms of specificity, as ACT limited the identification of wakefulness.
However, even so, these studies concluded that ACT is useful and valid for estimating total sleep time (TST) and Wake After Sleep Onset (WASO) 9,10 .
Nevertheless, in individuals with insomnia a moderate correlation was found between ACT and PSG 9 . It is likely that the reduced validity attributed to ACT in people with insomnia is due to its difficult in detecting periods of wakefulness (low sensitivity), which are higher in these individuals 9,11,12 .
However, because individuals with insomnia present high variability in sleep patterns, gauging with ACT for several consecutive nights has proved to be quite useful 13  Compared to PSG for evaluating sleep, in addition to being less expensive, ACT has advantages that have made it attractive to sleep researchers and physicians. This method can represent a useful tool to assess sleep for several consecutive days, having already been used to compare sleep parameters in different subgroups of patients.
Therefore, the purpose of this review is to evaluate the effectivity of ACT compared to PSG in determining sleep patterns of adults with insomnia.

METHOD
This study was developed as a systematic review and meta-analysis to evaluate the effectivity of ACT, having PSG as the gold standard exam. It was characterized as a study of the head-to-head type for effectivity's measurement. The

Research question
The research question was based on the acronym PICO -Population: adults of both sexes with insomnia; Intervention: ACT exam; Comparator: PSG; Outcomes: SL, TST, WASO, and SE. Thus, the final PICO question was: Is ACT as reliable as PSG to identify sleep patterns in adults with insomnia?

Research strategy
The search was made in the following databases on

Assessment of the quality of evidence
The GRADE system recommendations were used to evaluate the quality of the evidence, its classification (as high, moderate, low or very low) and its development 14 .
For this assessment, the evidence was separated into RCT and "observational studies". Following this approach, the RCTs started the evaluation with "high quality" and the observational studies with a "low quality" evidence classification. Then, the evidence was evaluated in relation to the five items that could alter its quality: the risk of bias, inconsistency, imprecision, the indirect evidence, and the publication bias 14 .

Data analysis
For the statistical analysis of the ACT compared to PSG the following variables were considered: TST, SL, WASO, and SE.
The meta-analysis was performed using the inverse of variance. The fixed effects model was used too and as the measure of the effect size, it was used the difference between averages. These were calculated and presented together with 95% confidence intervals (95% CI).
The meta-analysis was carried out using the Review manager 5.3, provided by Cochrane Collaboration. And Chisquare tests were calculated to estimate the heterogeneity.

RESULTS
The results found on the databases were reunited in a reference manager, from which 445 articles were retrieved.
The analysis and judgment of the documents were made by two independent judges, who decided from reading the papers' title and summary which ones should be included, considering the eligibility criteria. Duplicate studies were eliminated, leaving 386 articles to be analyzed. Then the eligibility criteria were applied, and 44 conflicts arose that were resolved by a third judge, resting 13 articles for complete reading by the end of the process. The entire inclusion and exclusion process considered the proposed steps by PRISMA FLOW 15 , which can be seen in Figure 1.

Main characteristics of the included studies
The following information was extracted from the manuscripts: reference, design, population, intervention, outcome, results, and limitations described in Figure 2. The six studies included evaluated the performance of ACT in comparison with PSG (head-to-head) for the outcomes covered in this study, namely: three observational studies and three randomized controlled clinical trials.

Metanalysis and interpretation of clinical significance
The meta-analysis was carried out with the six included studies, gathering data for each result of interest in the PICO question. All meta-analysis was performed using the fixed effects model, with the results displayed in a forest plot.
Below, the summaries of the identified evidence and the statistical analyzes performed are detailed.

Total Sleep Time (TST)
The meta-analysis of the six studies compared ACT with PSG for the assessment of TST in patients with insomnia. See level.

Sleep Latency (SL)
In the analysis for the evaluation of SL, on Figure 4, it suggests that there is no significant difference in the measurement of SL when using ACT and PSG, as it showed an average of -1.46 (95% CI: -9.61 on 6.70 min). This small difference indicates that ACT and PSG provide similar information to assess SL. The quality of the evidence was downgraded due to the imprecision and the heterogeneity was significant with an 85% level.

Wake After Sleep Onset (WASO)
The meta-analysis of the studies that compared ACT to PSG for the assessment of WASO, detailed on Figure 5, showed a confidence interval with a possible mean of -32.61 minutes (95% CI: -34.98 on -30.24 min). As a result, WASO was underestimated by ACT so ACT and PSG provide different measurements of this parameter. The quality of the evidence was moderate, and the heterogeneity was significant with a 94% level.

Sleep Efficiency (SE)
The analysis comparing actigraphy with PSG to assess SE, on Figure 6, showed a clinically insignificant range, with an average of 3.66% (95% CI: 3.02 on 4.30 min). This range indicates that ACT underestimated WASO compared to PSG.
The quality of the evidence was moderate, and the heterogeneity was significant with an 81% level.

Evidence quality
On Tables 1 and 2 are show the quality of the evidence for each outcome. Quality was classified as "low" due to inaccuracy (wide confidence interval); and "Moderate", due to the risk of bias (some studies presented other comorbidities concomitant with insomnia; they inserted subjective reports in the actigraphy; or they used ACT for several days and PSG only for one night) and inconsistency (high heterogeneity).    and in WASO (-43.56±5.72 min) were significant (p<0.0001 | mean discrepancy>0) and, with the exception of SL (p=0.1657) and SE (p=0.0746), the discrepancies were present in both age groups. The average discrepancies were higher in the older participants than in the younger ones for all sleep measures, except for SL (p=0.9099).        CI: Confidence interval. A. Patients had other comorbidities other than insomnia; B. heterogeneity greater than 50% was found; C. the studies compared the two interventions (head-to-head); D. in studies, actigraphy was used for several days and PSG only for one night; E. wide confidence interval; F. subjective report used.  with high evidence quality, found an average difference of 10.14 min between the two technologies, demonstrating that ACT can be safely used to assess the TTS of these individuals with insomnia 16 . Another study, however carried out with healthy adults, when comparing ACT with PSG to assess TTS, also found high reliability to be used in the objective measurement of this parameter 10 .

Revisão Sistemática
When evaluating SL, we found a narrow confidence interval (95% CI: -9.61 to 6.70 min) with an average difference of -1.46 min. Thus, ACT provides SL information that is true enough to be used as the only objective measure of sleep. Another study, when meta-analyzing 12 manuscripts to assess SL in a patient with insomnia, also found a narrow difference between ACT and PSG, with an average of 6.17 min agreeing with our study when stating that ACT can be used safely in the measure of this parameter 16 .
In the meta-analysis applied to WASO it was found an underestimation of ACT in relation to PSG in the measurement of this sleep parameter, with an average of -32.61 min. Therefore, based on this finding, the ACT did not show clinical confidence for this outcome. Another study, also in its analysis, found a clinically extensive interval between the two technologies, with possible average differences of 33.22 min, not indicating ACT as reliable as WASO to provide an objective assessment 16 .
The findings of this study were corroborated by a wide range of difference found between PSG and ACT for assessing SE, a 1% level, with the quality of its evidence being considered moderate 16 . In the present study, it was also found low reliability in measuring of the sleep by ACT compared to PSG, with an average difference of 3.66%. A different study, carried out with healthy population, also found actigraphy measures of SE different from the PSG's 10 .
ACT when compared to PSG has several benefits, such as: not requiring a sleep technician to use the device; easy recording of sleep for several consecutive days; low costs; it can be used in the individual's natural environment and it has greater ability to obtain reliable estimates of sleep parameters when patients may not be able to report them.
Skin irritation is the only damage reported from using the actigraphy but is considered mild when assessing its potential benefits 17 .
The main limitation of this study was the quality of the evidence from the studies selected for this review, in addition to the high heterogeneity found in the meta-analysis.
Besides, a tendency to most compare only one night of PSG records against several nights recorded with ACT was noticed, which can lead to greater bias associated with the effect of the first night due to the intrusive nature of the medication used in the PSG.
Other limitations found were the use of personalized measures of ACT and the fact that patients have comorbidities other than insomnia in some studies, which can cause information bias and heterogeneity. These limitations downgraded the quality of the evidence to moderate and low.

CONCLUSION
Based on this study, we concluded that ACT can be useful in assessing SL in patients with insomnia, providing a consistent measure with PSG. Although ACT showed differences in the assessment of TTS, WASO, and SE, the benefits of its use must be taken into account.
However, to affirm that these technologies are equivalent in terms of effectivity, more research is needed.
Understanding the differences between the two tools and addressing specific factors that may impact these