A scoping review of the evidence for the impact of pharmacological and non-pharmacological interventions on shift work related sleep disturbance in an occupational setting

Background: Shift work is essential in society but can be detrimental to health and quality of life and is associated with decreased productivity and increased risk of accidents. Interventions to reduce these consequences are needed, but the extent and range of trial evidence for interventions for those most affected by their shift-work schedules is unclear. We therefore carried out a scoping review to assess the availability of evidence to inform the development and evaluation of future interventions. Methods: We aimed to identify clinical trials of any intervention for shift work-related sleep disturbance that included a comparator group, where the intervention was delivered in an occupational setting. We searched Cochrane Central Register of Controlled Trials, Cochrane Database of Systematic Reviews, CINAHL, EMBASE, Medline and Science Citation Index from inception to 30 th March 2020 for relevant citations. Citations were screened by two independent reviewers, a third reviewer resolved disagreements. Data were extracted by two independent reviewers. Results: From 1250 unique citations, 14 studies met inclusion criteria for comparative trials of treatment in an occupational setting. There were five trials of hypnotics, five trials of stimulants, and four trials of non-pharmacological therapies (cognitive behavioural therapy, light therapy, aromatherapy and herbal medicine). Outcomes included sleep parameters, day-time sleepiness, and quality of life. There were no consistently reported outcomes across trials. Conclusions: Interventions fell into three distinct groups investigated in distinct time periods without progression from efficacy trials to wider-scale interventions. The lack of consistent patient-reported outcome measures limits synthesising findings. Some trials focussed on optimising sleep, others on reducing wake-time sleepiness. Adequately powered trials of existing interventions are needed, with the development and testing of novel combination treatments in patients with well-defined shift work sleep disorder. A core set of clinically relevant outcomes will develop and standardise the evidence-base for shift work sleep disorder.


Introduction
Shift work is often unavoidable in modern society. In 2015, 21% of the European Union working population carried out shift work 1 . Simplistically, shift work is a work schedule organised such that teams cover more than the usual 8 hour working day. More specifically it has previously been defined as "a work schedule in which at least 25% of work days involve the majority of working hours outside the time period between 08:00 and 17:00" 2 . For individuals, shift work has been linked to reduced quality of life, increased risk of workplace accidents 3,4 , sleep loss, obesity, type 2 diabetes, coronary heart disease, some cancers 5 and depression 6 . In an occupational setting this has been associated with lost productivity and increased errors 4 . This is of particular concern where work performance is related to health and safety, for example in healthcare, emergency response and mining industries.
Shift work sleep disorder (SWD) refers to the sleep disturbance experienced by a subset of shift workers who respond particularly poorly to their shift work schedule 7 . SWD is chiefly defined as having insomnia and/or excessive sleepiness temporally associated with shift work lasting ≥3 months, and where the sleep disturbance is not better explained by another diagnosis 8 . A diagnosis of SWD is associated with a greater risk of poor quality sleep, subjective health complaints and poor coping 9 , peptic ulcers, sleepiness-related accidents, absenteeism and depression when compared to shift workers without SWD 3 . SWD represents an area of therapeutic interest for individuals and institutions alike. However, the range of and evidence for interventions in mitigating associated impairments is unclear.
Existing reviews of interventions to improve sleep, sleepiness and related outcomes for shift workers found low quality evidence for some interventions 2,10 . Melatonin, armodafinil, modafinil, caffeine and naps were all found to have low-quality evidence for their efficacy in improving one or more outcome domains 2 .
Another review found studies of bright light, napping, physical exercise and sleep education as interventions for shift-workers, but concluded there was too much uncertainty to determine their impact 10 . To our knowledge, there has been no systematic review of clinical trials encompassing the full range of interventions focusing on shift-work related sleep disruption.
Here we present a scoping review of the available evidence from comparative studies examining the impact of interventions for SWD. We set out to identify all trials of shift workers with SWD, or sleep disturbance likely to be SWD that had a comparator group. We sought to establish the types and extent of available outcome data on this topic; how this has been reported and whether an informative quantitative data synthesis would be possible.

Study design
We carried out a scoping review to identify the main sources and extent of evidence available in the published literature 11 .

Eligibility criteria
Type of trials. We included randomised controlled trials (RCTs), randomised crossover trials and parallel group trials.

Population.
We included trials carried out with workers who were undertaking shift work and who had SWD as defined by the International Classification of Sleep Disorders (ICSD) criteria at the time when the trial in question was carried out. We also included studies where shift-workers were selected for having some level of sleep disturbance, that was not better explained by a known, non-SWD diagnosis, such as obstructive sleep apnoea (OSA) or narcolepsy. We excluded trials conducted on shift-workers unselected for having sleep problems, or conducted solely on shift-workers selected for having OSA or narcolepsy. We excluded studies in which airline cabin-crew or military personnel were the primary population group, as we considered the aetiology of SWD alongside frequent crossing of time zones likely to be different to that of SWD in the general population 2 .
We included only trials where shift workers received a study treatment in an occupational setting. We included trials where final outcomes were measured in a simulated shift work environment, if participants had been undergoing the treatment in a real shift-work environment previously. For example, where participants had taken modafinil or armodafinil before shifts for a number of months, and then completed a study night in a laboratory setting 12 .
Interventions. We included trials with any intervention, or combination of interventions, aimed at preventing or reducing the effects of SWD on sleepiness when awake, sleep disturbance, and associated functional impairment (e.g. reduction in wellbeing, depressive symptoms). We categorised interventions into pharmacological hypnotics, pharmacological stimulants, and non-pharmacological therapies. We included trials where

Amendments from Version 1
The most significant change in the new version of this article is the inclusion of a further paper in the review. This was a clustered randomised controlled trial of a combination intervention targeting shift-workers identified as high risk for shift work disorder. The intervention comprised coaching sessions, caffeine/melatonin and workplace lighting. Adding this paper did not change our overall conclusions that there is a general paucity of evidence for any intervention/s for shift work disorder.
Other changes include highlighting that one of the studied interventions (Brotizolam) is not actually licensed for use in the UK, US or Canada and another studied intervention (Shimian granules) are not mentioned in PubMed other than in the article we make reference to. If anything, this furthers our conclusions that studies of pragmatic interventions in this area are broadly lacking.

REVISED
interventions were compared to placebo, 'usual care', no intervention or to each other. Example interventions for pharmacological stimulants might include modafinil or armodafinil. Example interventions for pharmacological hypnotics might include melatonin or zopiclone.

Outcomes
We set out to identify and record when studies had used the following types of outcome measures. We did not define the instrument of measurement to be used for any given parameter.
We also planned to record when studies reported outcomes of: • Adverse events • Injuries or accidents whilst at work or commuting.

Search method
Electronic searches were conducted for Cochrane Central Register of Controlled Trials, Cochrane Database of Systematic Reviews, CINAHL, EMBASE, Medline and Science Citation Index from inception to 30 th March 2020. An example search strategy is shown in Table 1. The same search terms were used for all databases. We did not restrict by language. Title and abstract of all citations identified using this search strategy were screened for exclusion by two independent reviewers (R.C-J., E.D.) using Rayyan software 13 . Conflicts were resolved with a third reviewer (A.F.). The full text of the remaining citations was then reviewed. Citations were only included in the study where they referred to a published journal article. Therefore, drug-company listed trials without an associated journal publication were not included 14 .
Data extraction, analysis and synthesis Data were extracted manually by two independent reviewers using a custom spreadsheet [Microsoft Excel, Version 16.37, Copyright 2020]. We extracted study characteristics. We aimed to extract data about all outcomes relevant to this review to establish the extent of available data. Where studies reported multiple measures in the same domain (for example, multiple measures of quality of life), we selected externally validated measures.

Results
Search results and study characteristics From 1250 unique citations, 14 trials met inclusion criteria for data extraction and analysis. Two included citations reported different outcomes from the same trial 15,16 . As each citation reported different outcome data the trial is referenced to both citations 15,16 . Figure 1 presents a Preferred Reporting Items for 9 5 or 6 or 7 or 8 10 4 and 9 11 limit 10 to ("systematic review" or systematic reviews as topic or "reviews (maximizes specificity)") 12 randomized controlled trial.pt.
21 (quasi* adj2 random*).ti,ab. Systematic Reviews and Meta-Analyses (PRISMA) flow diagram showing reasons for study exclusions. In total, eight of the original included trials were randomised controlled trials 12,15-22 . The remainder were cross-over 23-25 and a parallel group design 26-28 . On peer review of this paper a further clustered randomised controlled trial was highlighted by a reviewer 29 . It was published after our pre-defined search period. However, it otherwise met inclusion criteria and so has been included to add to the data covered by this review 29 .
The studies were carried out over a wide geographical area, including Europe, Asia, and North America ( Funding came from a variety of sources, although those studies evaluating stimulants were funded by pharmaceutical companies. All studies whose authors declared a potential conflict of interest with a relevant pharmaceutical company were studies of stimulants 12,15,16,21,27,28 .    Table 3 shows the outcomes reported by the studies included in this review; the most commonly reported outcomes were TST and SOL, though they were measured by a variety of tools (polysomnography, actigraphy and sleep diaries).

Adverse events and drop-outs.
Adverse events were only reported by pharmacological studies. All studies of stimulants described monitoring for adverse events throughout the study periods. The protocols described varied but included combinations of subjective symptom reporting, physical examination and bedside and laboratory investigations. One study of zopiclone relied on study participants spontaneously reporting symptoms that they felt might be related to the study intervention 22 . Few studies reported total counts of participants affected by any adverse events. Some, but not all, studies reported where drop-outs were due to adverse events. Only one study explicitly defined how adverse event severity was categorised 12 and one other stated only that event severity was determined by a site investigator 27 .

Discussion
Our scoping review provides evidence that a review to assess the effectiveness of interventions to treat SWD would not provide sufficient data for a comprehensive meta-analysis. There are too few studies amongst shift workers with SWD for any particular intervention or category of intervention. Even where there is more than one study, there are often methodological limitations preventing pooling of data using meta-analysis, such as different measurement tools for one outcome. At least one of the hypnotic medications studied is no longer licensed in the UK, US or Canada 24 and one non-pharmacological supplement 17 is not described elsewhere in published literature. This limits the practical usage of these studies. Whilst we considered that a narrative review might be possible, the represented methodologies, interventions and data were so heterogenous it was not deemed appropriate at this point.
We did not critically appraise the included sources of evidence as this scoping review was conducted to provide an overview of the existing evidence regardless of methodological quality or risk of bias. However, we note that there were high rates of participant attrition in experimental and control arms in both pharmacological and non-pharmacological studies. Many drop-outs were unexplained.
Previous systematic reviews looking at a wider population of all shift-workers have similarly found a paucity of evidence regarding impact of interventions 2,10 . There are many interventions that have been considered in the literature for sleep disturbance 30,31 that do not appear to have been evaluated in clinical trials with populations that meet the criteria for SWD, for example individual level interventions such as napping 32 and institutional level interventions such as optimising shift schedules 33 . These interventions could be considered in future work looking at the SWD population specifically.
In this scoping review, we have grouped the short-term trial outcomes into sleep outcomes (ease of falling asleep, the continuity, duration and quality of obtained sleep), wake-time outcomes (sleepiness during waking hours), combined sleepwake outcomes and other aspects of impairment. Sleep outcomes are the primary target of hypnotic pharmacological interventions whilst wake-time outcomes are that of wakefulness-promoting (stimulant) pharmacological agents. Non-pharmacological interventions vary in their putative mode of action.
The lack of consistent measurement of sleep outcomes was perhaps unsurprising then. Studies of the efficacy of stimulant pharmacological agents were aiming to promote wakefulness whilst at work and so often used real time state-based assessments like KSS and SSS. These were not used by any study of hypnotics or non-pharmacological interventions. Multiple studies reported a range of sleep outcomes (TST, SOL) but instrument of measurement was variable. Further work in this area could establish useful and replicable outcome measures for SWD.
The MSLT is described as the gold standard measure of sleepiness and is objective but many of the studies included in this review only used subjective measures of sleepiness 15,16,21 . Whilst subjective measures of sleepiness are clinically relevant and sensitive to insufficient sleep acutely 34 , they have a non-linear relationship to sleep debt and correlate poorly with sleep debt in the context of chronic sleep deprivation 35 .

CGI-C HADS AE
Sadeghniiat- Further, occupational impairment in SWD has been shown to be more strongly correlated to insomnia than to sleepiness so focusing on sleepiness may be unhelpful for work related outcomes 36 . As such, it is likely a useful set of outcome measures for SWD will cover multiple domains of the disorder: insomnia symptoms, excessive waketime sleepiness and impact on functioning. These need to be consistent and standardised to prevent research waste where data from one study cannot be used to inform subsequent work or contribute to systematic reviews of the problem. We envisage shift-worker participation in future outcome measure development will be beneficial.
Advice for shift workers using available evidence has been described pragmatically elsewhere 37 . There are also multiple guidelines available for clinicians treating patients with sleep disturbance due to shift work 38,39 . The evidence for the effectiveness of interventions currently available is unclear and the lack of follow-on evaluations from those identified in this review confirms the continuing uncertainty and gap in the evidence. Novel treatments and combinations of current treatment, targeting individuals where the type of shift work and the combination of symptoms experienced by an individual are needed.

Data availability
All data underlying the results are available as part of the article and no additional source data are required.

Ulises Jiménez Correa
Universidad Nacional Autonoma de Mexico, Mexico City, Mexico City, Mexico It is a very interesting scoping review with the aim to describe the evidence for the impact of pharmacological and non-pharmacological interventions on shift work disorder. Beyond interesting, I consider that it review is necessary.
I believe that the strength of this review consists in highlighting the little research that aims to determine the evidence of pharmacological/psychological treatment for sleep disorder due to work hours. The issue is very important because we live in a 24-hour society (i.e. factories, public security, mobility, hospital services), that require of workers for night shift and shift rotation; those who are at high risk of suffering symptoms of insomnia and daytime sleepiness; as well as metabolic alterations related to circadian rhythm disorders and accidents.
Methods are according to introduction; results with the aim; and discussion with the results. However, with the aim of complementing the results section with the necessary information, I think that it is necessary to include an extra column in table 3 with the main findings of the included studies.
I do not have any more comments.

Are the rationale for, and objectives of, the Systematic Review clearly stated? Yes
Are sufficient details of the methods and analysis provided to allow replication by others? Yes

Is the statistical analysis and its interpretation appropriate? Not applicable
Are the conclusions drawn adequately supported by the results presented in the review?

Yes
The paper is clear, well written and comes from expert colleagues. However, this is essentially stating that there is insufficient evidence to carry out meta-analysis -I agree with this and that there is very limited data. In fact there are no two trials using the same methodology or intervention. A minor point but I disagree with the opening sentence, there isn't really any evidence for shift work increasing, most work fewer hours with more protection of sleep periods in legislation compared to any period from the industrial revolution onwards. About 20% of people work some form of shift but this figure has been consistent over some time. Therefore many people work shifts, but always have. There isn't a clear definition of what a shift is or what makes a shift worker to help the non-expert.
I had some questions about the interventions that were included and discussed. Brotizolam is a potent, old hypnotic no longer licensed for use in the UK, US or Canada which I think either needs to be clearly stated in the text or this study not used for those reasons. Likewise -there is a single PubMed reference only to shimian granules -a product that is not licensed and there is simply no information at all on what they are -a chinese complimentary medicine not in any way standard for sleep clinics. Again, there is a single RCT but nothing else at all in PubMed relevant for shimian granules.
I appreciate the authors decided against critical review or narrative review but a small amount of extra information still seems necessary here.
Accepting the timelines of any systematic review and inevitably -more studies will follow. However given such limited data -there is a very recent RCT using CBTi strategies just published 1 . I think the paper ideally should be revised and this paper should be included, there is so little data that this would be a valuable addition, no stats were possible and all information presented is simply descriptive so I do not see this taking the authors significant time.
There is also another drug company-listed trial with all results in clinicaltrials. Revision 5 "Accepting the timelines of any systematic review and inevitably -more studies will follow. However given such limited data -there is a very recent RCT using CBTi strategies just published 1 . I think the paper ideally should be revised and this paper should be included, there is so little data that this would be a valuable addition, no stats were possible and all information presented is simply descriptive so I do not see this taking the authors significant time." ○ Yes thank you so much for raising this. We agree this is a useful trial to highlight given the paucity of trials otherwise. We have simply included it and explained in the Results section how it has been added after peer review (first paragraph of Results). Whilst this isn't perfect we agree that including it, even post-hoc, is in keeping with the goals of a scoping review. It is a valuable addition for completeness of this paper though it has not changed our conclusions ○ The main edits pertaining to this are found in Figure 1, Tables 2 and 3 and within