Effectiveness of internet-delivered cognitive behavioral therapy for insomnia: a systematic review and meta-analysis

ABSTRACT This systematic review and meta-analysis aimed to assess the effectiveness of internet-delivered cognitive behavioral therapy for insomnia (ICBT-I) in the short and long term. We searched five electronic databases for studies published from June 2000 to May 2022. Our analysis included 27 studies with 8,728 participants, of which 4394 were in the experimental group and 4334 were in the control group. ICBT-I significantly improved the sleep scores of the experimental group by reducing the intensity of insomnia (ISI) (−4.32) in the post-intervention (I2 = 94%) (95% CI −6.15 to −2.48, p < 0.0001, g = 0.760) and follow-up phases (−2.92) (I2 = 95%) (95% CI −4.87 to −0.97, p < 0.0003, g = 0.622). This therapy also effectively improves short-term and long-term insomnia symptoms, including sleep onset latency, wake after sleep onset, night awakenings, sleep efficiency, and total sleep time. That holds the potential for extensive utilization across diverse populations suffering from insomnia, surpassing the limitations associated with conventional in-person therapeutic methods.


Introduction
On average, humans spend about one-third of their lives in sleep (Ohayon 2011).Sleep is an essential part of survival, which plays a crucial role in biological and psychological regeneration processes.Sleep deprivation increases numerous physical and psychological risks (Liu et al. 2017).Insomnia is a prevalent disorder affecting a large proportion of the population and is a significant clinical concern for mental health practitioners (Winkelman 2015).Insomnia has symptoms such as difficulty starting and maintaining sleep and waking up in the morning (Riemann et al. 2015).Approximately 30%-35% of adults report insomnia symptoms, and at least 10%-15% of people chronically suffer from this disorder (Riemann et al. 2017).Studies show that persistent insomnia is significantly associated with various physical and psychological problems, such as the increased likelihood of suicide, alcohol and drug abuse, anxiety, depression, and heart failure (Taylor et al. 2005;Grandner et al. 2012).In addition, it imposes many social costs on patients and society, such as increased use of health services, absenteeism or reduced productivity at work (Wickwire et al. 2016), and reduced daily performance (Jansson-Fröjmark 2014).Therefore, the treatment of this problem is essential.Although drug therapy is the most commonly known treatment method in the treatment of insomnia and reduces its symptoms (Cho and Song 2014), over time, sleeping drugs such as benzodiazepine receptor agonists have side effects such as headache, dysfunction, dependence, and tolerance, and they cause complexity in the treatment process of this disease (Riemann and Perlis 2009).For this reason, psychological interventions for insomnia to identify and manage sleep problems have received the most attention in studies.Cognitive behavioral interventions are the first-line and gold-standard treatment for insomnia.Several studies have demonstrated this type of intervention to be a multifaceted treatment with minimal complications (Cheng et al. 2019), which is more successful in treating acute insomnia and producing long-term results (Mitchell et al. 2012).Furthermore, its impact on primary insomnia and insomnia with various medical and psychiatric conditions was well-proven (Ballesio et al. 2018).This evidence-based non-pharmacological treatment is a multi-component treatment that changes the sleep quality of the affected person by changing dysfunctional cognitions and behaviors (Thakral et al. 2020).The main goal of this treatment is to change the behavioral (improper sleeping habits, irregular sleep schedule), cognitive (unrealistic expectations, worry, unhelpful beliefs), and physiological (mental and physical tension, excessive arousal) factors that perpetuate insomnia.This treatment is usually provided in four to eight sessions and at weekly intervals.The number of visits can differ depending on the severity of insomnia, co-morbidities, and patient motivation (Morin and Benca 2012).Studies comparing the cognitive-behavioral treatment of insomnia with drug therapy also show the beneficial effects of this short-term treatment model and the stability of these effects over time (Muench et al. 2022).Despite the effectiveness and efficiency of approaches such as cognitive-behavioral psychotherapy for insomnia, various studies have indicated the limitations of using this treatment (Shaffer et al. 2021) and several obstacles such as limited access to trained therapists, relatively high costs of faceto-face treatment (Espie et al. 2012), time limitations, travel costs, and the fear of the mental illness stigma has been reported.These are the barriers to making this treatment accessible, affordable, and widely available for the population suffering from insomnia.They also highlight the need to plan easily access this treatment (Savard et al. 2021).One of the newest ways to overcome these challenges in recent years is to present them in an intelligent, automatic, semi-automatic way using computers and smartphone applications and on the web, known as internet-delivered interventions (Titzler et al. 2020).This set of treatments began to develop based on the internet in the late 1990s to facilitate access to mental health services for all society groups (Ebert et al. 2015).Nowadays, they utilize for a wide range of disorders (Schlarb et al. 2020).This treatment is a combination of webbased self-help interventions.This method provides diverse educational content such as short articles, animations, short clips of the process of sleep formation, teaching the principles of sleep, and video communication with the therapist if necessary (Dever Fitzgerald et al. 2010).In 2004, a study published the first randomized controlled trial evaluating an Internet-delivered intervention for insomnia (Ström et al. 2004).Based on the guidelines for cognitive-behavioral therapy of insomnia, this treatment included sleep hygiene education, sleep restriction, stimulus control, cognitive therapy, and relaxation techniques.Studies have shown its beneficial effects on the total amount of sleep time, wakefulness after sleep onset, and sleep efficiency (Zachariae et al. 2016).Following this study, several researchers, such as Zachariae et al. (Zachariae et al. 2016) andYe et al. (Ye et al. 2016), confirmed the effectiveness of this treatment model.However, the results of some controlled studies indicate that each of these interventions has unique effects on insomnia symptoms and was effective to a certain extent in treating this disorder.In addition, there is a need to investigate the long-term effectiveness of this type of intervention (Lorenz et al. 2019).Therefore, due to the increasing use of Internetdelivered interventions for insomnia treatment, the increase in the number of clinical trials of internet-delivered cognitive behavioral therapy for insomnia (ICBT-I), the lack of review of the effectiveness of these interventions in new studies, and also the limitation of meta-analyses conducted in this field, we intend to provide an update in this field and evaluate this type of intervention's short-and long-term effectiveness.In this study, we evaluated the effectiveness of indices such as insomnia severity, sleep onset latency (SOL), wake after sleep onset (WASO), the number of nighttime awakenings (NWAK), sleep efficiency (SE), and total sleep time (TST).The main goal of the current research is to answer these questions: Are internet-delivered cognitive -behavioral therapy (ICBT) effective on insomnia?Furthermore, if yes, what was the effectiveness of Internetdelivered psychotherapies on insomnia in the short and long term?

Method
We used the meta-analysis statistical technique based on the Preferred Reporting Items for Systematic Review and Meta-Analysis (PRISMA) statement to collect, combine, and summarize the research findings (Moher et al. 2009).Glass (1976) introduced meta-analysis (Cheng et al. 2019).The purpose of meta-analysis is to use techniques that lead to the integration and analysis of the pattern of findings obtained from different studies about a research question (Riemann and Perlis 2009).The statistical population of the present meta-analysis was all valid, available, and related studies on the effectiveness of internet-delivered interventions on adults with insomnia that met the necessary methodological conditions.The sample was 27 articles based on the inclusion and exclusion criteria.The method of conducting the research was as follows: in the first step, in order to obtain studies related to the purpose of our research, two authors searched the keywords of internet, internet base, web, web base, online, digital, self-help, self-administer, self-care, cognitive therapy, cognitive-behavioral therapy, and sleep problem or insomnia and their word-family using OR and AND operators, in the period from June 2000 to May 2022 in databases of PubMed, Scopus, Cochrane Central Registry of Controlled Trials (CENTRAL), PsycINFO, Embase systematically searched.Additional details regarding the search strategies are given in the supplementary file (Appendix-A).In the next step, after removing duplicate studies, we examined the studies obtained from the databases according to the inclusion and exclusion criteria.The inclusion criteria were: being in the field of internet-delivered interventions, adult participants (18 years and older) with a clinical diagnosis of insomnia according to the Diagnostic and Statistical Manual of Mental Disorders (DSM) or the International Classification of Sleep Disorders (who were categorized based on dimensions such as insomnia severity, sleep efficiency, the onset of sleep latency, awakenings after sleep onset and the total amount of sleep time and mental quality of sleep), having appropriate methodology (studies with appropriate study design, clear and well-defined objectives, robust data collection methods, and appropriate statistical analyses) and access to the full text of the studies.Moreover, the exclusion criteria were: duplicated studies, studies that did not have suitable methodological conditions and did not provide information about pretest-posttest scores, changes in groups, sample size, and effect size (such as Cohen's D and the eta coefficient), descriptive, review, letters to the editor, correlational and case studies, articles that focused on problems other than insomnia (fatigue, depression, and pain) or the existence of insomnia due to biological causes.Finally, the obtained studies were evaluated based on the content checklist.To determine the inclusion or exclusion of studies, after reviewing the titles and abstracts of the articles, we extracted their full text if they met the inclusion criteria.The study referred to the third author when the two authors disagreed.Then we read the full text of the articles having inclusion criteria and checked the exclusion criteria.After considering the entry and exit criteria, 103 studies were included.Finally, we selected and analyzed 27 studies after evaluation and final agreement between the reviewers regarding the eligibility of each study (Cohen's Kappa = 87%).

Content checklist
We used the content analysis checklist to select the research, check the inclusion and exclusion criteria, and extract data from included studies.This checklist consists of 12 items: the study title, first author, year of publication, sample characteristics, sample size, treatment components, average age, duration of treatment, follow-up, method of providing treatment, the amount of each statistic before and after the intervention, and the level of significance (Table 1).We evaluated the quality of the studies using sleep-relevant study quality criteria that were used in previous related studies (Zachariae et al. 2016), which included 12 quality criteria: 1) description of the selection process of the participants, 2) detailed description of the process of a random selection of participants in experimental and control groups, 3) description of the method of withdrawal and exit from the study, 4) detailed definition of the research subjects, 5) definition of the method of measuring the research result, 6) description of the entry and exit criteria 7) description of the reasons for choosing the sample size, 8) description of statistical methods, 9) report of findings, 10) definition of sleep and related problems, 11) description of sleep intervention components and level of involvement of participants, 12) description of the participants who were diagnosed with insomnia based on DSM-5 criteria (coded 0 for no, 1 for uncertain to some extent, or 2 for yes, and the total score was in the range of 0 to 24).We performed data analysis using comprehensive meta-analysis software.Thus, after completing the PRISMA checklist, we entered the obtained data into the comprehensive meta-analysis software (CMA3) and then calculated the effect size of the studies using this software based on Hedges and Olkin's (1985) formula.Hedges and Olkin (1985) defined the effect size as the difference between the average of the experimental and control groups, considering the standard deviation.

Results
After searching the databases, we obtained 1711 studies using the research keywords.After removing duplicates, 1627 studies remained.After the initial study of the titles and abstracts of the articles, we excluded 1524 studies due to the lack of inclusion criteria and selected 103 articles to review their full text.In the next step, we read the full text of 103 articles.Then, in terms of the exclusion criteria, we excluded 76 studies from the research process (Figure 1).Finally, we collected the primary data of the current research from a total of 27 experimental studies of internet-delivered psychotherapy on insomnia (Table 1).The total subjects of the studies were 8728, of which 4394 were in the experimental group and 4334 were in the control group.Five studies were conducted on patients with comorbid diseases such as cancer, cardiovascular diseases, chronic pain, depression, and anxiety.A total of 22 studies were conducted on adults with insomnia.In addition, 18 studies were conducted in Europe, one in China, one in Japan, and seven in the United States of America.All studies used the randomization method to place the subjects in the experimental and control groups, and no double-blind method was used in any of the studies.The minimum intervention time was four weeks and the maximum time was ten weeks, and the average total intervention time was 6.6 weeks.Among the studies, 18 were clinical trials, and nine were experimental.The waiting list group was sleep hygiene education in 4 studies, face-to-face sleep-based intervention in 3, and placebo in one.ICBT, in all studies, was a multi-part intervention, including behavioral, educational, and cognitive techniques used in traditional insomnia cognitive behavioral therapy.This treatment includes 1) controlling stimuli as instructions to strengthen the relationship between bed, bedtime, and sleep.2) Sleep health education that focuses on lifestyle and environmental factors related to sleep (such as avoiding caffeine and alcohol before sleep and increasing physical activity).3) cognitive therapy that aims to change unhelpful sleep-related thoughts (such as worrying about the consequences of insomnia) and beliefs (such as: having eight hours of sleep to maintain health).4) Sleep restriction to adjust the sleep and wake schedule and reduce the proportion of waking time spent in bed.5) relaxation techniques aimed at reducing anxiety and hyperarousal before sleep.The amount of interaction between the participant and the therapist was different, and 17 studies were completely automated, interactive, and personalized.In comparison, three studies were semi-directed by the therapist and included online and offline content about the process of sleep problems.The therapist encouraged the patient and checked the completion of their task.Seven treatments were also guided by the therapist and entirely online.
In order to measure the severity of sleep problems, studies have used a variety of tools.Eighteen studies reported the findings related to measuring the intensity of insomnia, including the Insomnia Severity Index (ISI), Sleep Condition Indicator (SCI), and Sleep-50 scale.Also, nine studies measured the sleep efficiency index.The secondary outcomes were sleep onset delay, total sleep time, subjective quality of sleep, awakening after sleep onset, the number of night awakenings, and total time in bed.The average duration of the treatment period was 6.6 weeks (between 4 and 10 weeks).In twelve studies, follow-up treatment was performed after the treatment to check the continuity of the results.Its duration was between 1 and 18 months, with an average of 7.5 months.

The effect of ICBT on insomnia components in the post-test phase
In the post-test phase, 23 studies reported the data related to the Insomnia Intensity Index (ISI) obtained from 3764 participants in the internet-delivered intervention group and 3796 participants in the control group.The data analysis showed that ICBT in the experimental group compared to the control group significantly improved the severity of insomnia and was associated with a decrease in ISI (−4.32) (I2 = 94%) (95% CI − 6.15 to −2.48, p < 0.0001, hedges effect size g = 0.760) (Figure 2).Moreover, the results of the studies indicated that ICBT-I on the dimensions of sleep onset latency (SOL); −15.09 min (95% CI −22.33 to −8.04, P < 0.0001) (I2 = 98%) wake after sleep onset (WASO); −20.37 min (95% CI −28.03 to −12.70, P < 0.0001) (I2 = 98%); the number of awakenings (NWAK); −0.28 times (95% CI −0.44 to −0.11, P < 0.0001) (I2 = 83%), sleep  2).The study's clinical implications were that participants who received ICBT for insomnia fell asleep an average of 15.9 minutes earlier, reported 20.37 minutes fewer night awakenings, and 16.52 minutes more total sleep time compared to the control group.They reported an average of 8.5% higher sleep efficiency.

Effect of ICBT on sleep at follow-up
We calculated the changes between the pre-test and follow-up assessment in ICBT groups.In our study, follow-up is defined as four weeks to 18 months after the end of the intervention.Examining the effect size among the 12 studies that have followed up showed that after completing the treatment, there was improvement and reduction (−2.92) in the insomnia intensity scores of the experimental group compared to the  control group (95% I = 2) (95% CI −4.87 to − 0.97, p < 0.0003, hedges effect size g = 0.622) remained in the follow-up phase (Figure 3).Moreover, the results showed that in ICBT-I, the dimensions of sleep onset delay (SOL); −14.26 min (95% CI −28.03 to −0.49, P < 0.042) (I2 = 98%) wake after sleep onset (WASO); −21.80 min (95% CI −51.64 to 8.02, P < 0.15) (I2 = 99%); the number of night awakenings (NWAK); −0.06 times (95% CI −0.24 to −0.11, P < 0.50) (I2 = 54%), sleep efficiency (SE); 6.42 (95% CI 1.38, 11.46, P < 0.001) (I2 = 98%), and total sleep time (TST); 11.50 min (95% CI 8.25, 17.21, P < 0.007) (I2 = 94%), had significant positive effects in the follow-up phase and caused the disease improved in the subjects of the experimental group compared to their counterparts in the waiting list group.In addition, compared to the pre-test, ICBT-I significantly affects the severity of insomnia, sleep onset delay, awakening after sleep onset, the number of night awakenings, sleep efficiency, and total sleep time during the seven-month follow-up (Figure 4).

Sensitivity analysis and publication bias
The risk of publication bias is a widespread problem when conducting metaanalyses, which can be tested and reported with funnel plots and Egger's test.In the present study, we performed a sensitivity analysis to investigate outlier data and the risk of publication bias to determine the outlier effect sizes using funnel plots.The funnel plots obtained for the Insomnia Intensity Index (ISI) in the postintervention phase show symmetry (Figure 2).The significant values of Egger's regression obtained in the post-intervention phase of insomnia intensity index (ISI) were 0.64 and sleep efficiency (SE) was 0.32, as well as Duval and Tweedie's Trim values in insomnia intensity index (ISI) g = 0.22 (95% CI: −1.16, − 0.97) and sleep efficiency (SE) g = 0.36 (95% CI: 6.12, 10.44) was obtained, which indicates the absence of publication bias.

Discussion
We conducted this systematic review and meta-analysis to investigate the effectiveness of ICBT-I.The obtained results indicated that ICBT-I as an available treatment significantly improved the components of insomnia.In addition, ICBT-I had significant positive effects on dimensions such as sleep efficiency, sleep latency, total sleep time, and post-sleep awakenings in the post-test phase.These findings show that this intervention improved sleep onset and retention in people with insomnia.People receiving this intervention fell asleep an average of 15.9 minutes earlier and reported 20.37 minutes fewer night awakenings.Moreover, the increase in sleep efficiency and total sleep time indicates that the experimental group experienced better quality sleep and reported an average of 16.52 minutes more sleep duration.In this study, we examined the follow-up effect size in twelve studies to check the continuity of the obtained results over time.The results showed that this treatment, in addition to having a good effect in the short term (1-18 months), also significantly improves subsequent follow-ups.The findings of this study, due to the large number of samples of the studied studies (N = 27) compared to the studies mentioned above and the long follow-up period (on average 7.5 months) in the meta-analysis, provide more substantial support than previous studies for the effect of ICBT on multiple dimensions of insomnia.Also, it shows that this type of available intervention can be widely used to reduce people's insomnia problems.Our study also found that similar to traditional methods in which therapy provides through face-to-face interactions with a psychologist (Bastien et al. 2004), ICBT-I participants can receive feedback during treatment and benefit from lower cost and higher efficiency (Okajima 2015).In addition, patients can receive intervention at any time and place to quickly learn their cognitive skills and behavioral strategies and communicate with their therapists (Ritterband and Thorndike 2012).Another result was that the studies used packs with relatively similar structures but different content.There were also differences among the results obtained from each intervention, so each program had its characteristics that created different levels of efficiency.One of the assumptions in internet-delivered interventions is that the amount of clinical support received by the patient is important to the effectiveness of the intervention provided by the internet (Andersson et al. 2009).According to the studies, receiving more clinical support by the patient was associated with a more significant effect size for sleep efficiency after treatment and insomnia severity in the follow-ups.One of the reasons for this result is the increased adherence to treatment following clinical support, which means that by receiving support, a person will show more adherence to treatment.In addition, some studies (Andersson et al. 2009;Ritterband et al. 2009) show that fully automated systems can be more effective because these interventions provide significantly more interactive and tailored components for patients.Another finding in line with Soh et al.'s (Soh et al. 2020) review of subgroups was that the existence of medical diseases and psychiatric disorders related to insomnia did not significantly affect the average difference between studies.In other words, this intervention can be used for various clinical and non-clinical populations.In addition, although the review by Zachariae et al. (Zachariae et al. 2016) indicated that interventions with a longer duration and with more personal guidance have a better result, these interventions caused the dropout of participants and resulted in a smaller effect size.However, among the related questions that have yet to be examined is the issue of cost.If personalized clinical support is considered essential to improve the effectiveness of these interventions, it is crucial to consider the cost of the treatment provided.In general, adding human support can significantly increase costs and reduce people's accessibility.In this way, comparing and determining who would benefit from intelligent and automatic interventions and who would benefit from semi-automatic interventions could help improve efficiency and reduce overall costs.
This updated meta-analysis had several strengths and weaknesses, like previous studies.The first strong point of this study was that the reviewed studies were among the most recently published articles and had high quality.Moreover, they all reported the required data to calculate the effect size.Also, in all studies, the ICBT-I components used were clearly defined, and their effects were described in the pre-test, post-test, and followup phases.This feature allowed the analyst to examine the combined effect size and the possible role of different moderators.Finally, in our study, it was possible to check the expectation bias and calculate the effect size of the studies according to these biases.
Despite these strengths, some factors can limit the interpretability of the results.The first factor was the moderate to considerable heterogeneity that existed among studies.These factors indicate systematic differences between studies and do not necessarily indicate random sampling error (Sterne et al. 2008).This heterogeneity among studies is not a defect or weakness.Heterogeneity in the included studies indicates the existence and analysis of moderating variables that the effect sizes of the studies are affected by the changes of these factors.These factors include the duration of the treatment, the dropout rate, and other characteristics of the therapy.Another limitation of the present study was the sample size in several studies, which might effectively affect the final effect size.We will need more confirmation, considering the missing studies and adjusting the effect size based on it.Also, most of the available studies were on patients in Europe, the United States, and the United Kingdom, and there were a few studies conducted and published in Asian countries.
Finally, due to the limited number of follow-up studies, this part of the study needs further investigation, and caution should be observed in interpreting the results of this part.As discussed earlier, future studies are needed to examine the role of potential moderators of ICBT-I effect size to evaluate the efficacy of this group of interventions.It is also necessary to compare face-to-face therapy with internet-delivered therapy to investigate and compare the effect size of each intervention.Moreover, by specifying the beneficial aspects of this type of intervention, it is possible to help their growth and expansion.In addition, the implementation of internet-delivered psychotherapies and the investigation of their cultural aspects on disorders such as insomnia is one of the research gaps in this field that needs further studies.We suggest that future studies investigate clinical trials comparing ICBT-I with face-to-face interventions and drug therapy and report more results in this field.Finally, due to the growing popularity of providing online treatment, this method can be used for other helpful self-help approaches.
Due to the increasing growth of internet use and the expansion of access to online mental health and healthcare services, many studies had focused on the effectiveness of designed interventions.The meta-analysis of recent clinical trials also shows that online treatments are effective on a wide range of disorders such as anxiety, depression, alcohol consumption, and smoking are effective (Glozier et al. 2019).Since many people with psychological injuries experience insomnia, which increases the severity of the disorder's symptoms, access to easy treatments such as internet-delivered interventions can prevent the aggravation and occurrence of subsequent problems.It was also found that this group of interventions provides more benefits by reducing access and transportation costs for people who do not have easy access to psychological services and providing access to professional interventions at a low cost (Morin 1993).Moreover, since the stigma of mental illness makes some patients unwilling to receive treatment, the existence of online intervention programs makes the person receive services with more focus by reducing anxiety and fear caused by the stigma of mental illness.

Conclusion
In general, the results of the present meta-analysis on all the studies conducted in the field of ICBT-I showed statistically significant effects of this type of intervention on the severity of insomnia and various sleep-related outcomes.We can compare this intervention with face-to-face interventions and drug therapy.This study showed that this type of intervention is beneficial and expanding and have significant effects on sleep dimensions and components such as sleep efficiency, sleep latency, total sleep time, and awakenings after sleep onset of the participants.Also, our study demonstrated that ICBT-I has wide applicability for various sleep disorders.Although few studies had follow-ups after the completion of the therapy, the results obtained from the analysis of the follow-up data showed that the effects of this therapy had maintained over time.

Disclosure statement
No potential conflict of interest was reported by the authors.

Funding
We have no affiliations with or involvement in any organization or entity with any financial interest or non-financial interest in the subject matter or materials discussed in this manuscript.

Figure 1 .
Figure 1.PRISMA flowchart to illustrate the process of research and inclusion of studies in the systematic review.

Figure 3 .
Figure 3. Mean differences in ISI between internet-delivered cognitive behavioral therapy for insomnia and wait list in post-test.
These updated results are in line with similar studies such as Zachariae et al. (Zachariae et al. 2016), Ye et al. (Ye et al. 2016), Seyffert et al. (Seyffert et al. 2016), and Soh et al. (Soh et al. 2020) showed a decrease in the intensity of insomnia of people receiving ICBT.

Figure 4 .
Figure 4. Mean differences in ISI between internet-delivered cognitive behavioral therapy for insomnia and wait list after follow-up.

Table 1 .
Summary of characteristics of studies eligible for inclusion criteria.

Table 2 .
Summary of the pooled effect sizes for primary and secondary outcomes.
b N the total number of participants in ICBT-I and control.c MD refers to mean difference.d Statistically significant results shown in bold.