The Clinical Effectiveness of Cognitive Behavioral Therapy for Patients with Insomnia and Depression: A Systematic Review and Meta-Analysis

Background Insomnia and depression often co-occurr. However, there is lack of effective treatment for such comorbidity. CBT-I has been recommended as the first-line treatment for insomnia; whether it is also effective for comorbidity of insomnia and depression is still unknown. Therefore, we conducted this meta-analysis of randomized controlled trials to assess the clinical effectiveness and safety of CBT-I for insomnia comorbid with depression. Data Sources. Seven electronic databases, including China National Knowledge Infrastructure (CNKI), Wanfang Database, China Science Technology Journal Database, SinoMed Database, PubMed, the Cochrane Library, and EMBASE, as well as grey literature, were searched from the beginning of each database to July 1, 2019. Study Eligibility Criteria. Randomized controlled trials that compared CBT-I to no treatment or hypnotics (zopiclone, estazolam, and benzodiazepine agonist) for insomnia comorbid with depression and reported both insomnia scales and depression scales. Study Assessment and Synthesis Methods. Cochrane Reviewer's Handbook was used for evaluating the risk of bias of included studies. Review Manager 5.3 software was used for meta-analysis. Online GRADEpro was used to assess the quality of evidence. Results The pooled data showed that CBT-I was superior to no treatment for insomnia, while it was unsure whether CBT-I was better than no treatment for depression. And the effectiveness of CBT-I was comparable to hypnotics for both insomnia and depression. CBT-I was likely to be safe due to its noninvasive nature. The methodological quality varied across these trials. The evidence quality varied from moderate to very low, and the recommendation level was low. Conclusions Currently, findings support that CBT-I seems to be effective and safe for insomnia comorbid with depression to improve the insomnia condition, while it is unsure whether CBT-I could improve depression condition. More rigorous trials are needed to confirm our findings.


Introduction
Insomnia is a kind of sleep disorder, and people with insomnia are unsatisfied with their sleep time and sleep quality. Insomnia patients have difficulty initiating, maintaining sleep or returning to sleep [1]. Several studies have shown that 6% to 10% of adults suffered from insomnia [2][3][4], and 10-15% of insomnia patients tended to develop into chronic course [5]. Insomnia increases the risk of many health problems including suicidal ideation and behavior [6], cardiovascular diseases [7], depressive disorder, arterial hypertension, myocardial infarction [8], chronic heart failure [7,9], type 2 diabetes [10], and cognitive impairment [11], which would place a heavy burden on society and individuals. Treatments for insomnia included cognitive behavioral therapy, pharmacologic therapy, and complementary and alternative therapy. Pharmacologic therapy for insomnia includes benzodiazepines (triazolam, estazolam, temazepam, flurazepam, and quazepam), nonbenzodiazepine hypnotics (zaleplon, zolpidem, and eszopiclone). Complementary and alternative therapy for insomnia includes acupuncture and Chinese herbal medicine. Moderate-quality evidence showed that CBT-I improved sleep outcomes in the general population, including reduced sleep onset latency and wake after sleep onset, and improved sleep efficiency and sleep quality. And any harm associated with CBT-I is likely to be mild, so CBT-I was recommended as the first-line treatment by the ACP Clinical Guidelines Committee [12,13].
In fact, comorbid insomnia is commonly seen in clinical practice. And insomnia always combines with depression. Symptoms of depression include depressive mood, decreased interest, lack of motivation, and fatigue [14]. Insomnia and depression are often influenced by each other, and this mutual influence may increase the risk of suicide. A recent study suggests that sleep problems are associated with severe depression, suicidality, and worse outcomes for treatment of depression [15]. And another study shows that early changes in insomnia characteristics may predict longterm depression outcomes [16]. No therapy has been proven to be effective and safe for this comorbidity at present. Sequential approach and concomitant treatment for such comorbidity are both in their preliminary stages; welldesigned randomized controlled trials with long-term follow-up are needed to evaluate the effectiveness and safety of these treatments.
Cognitive behavioral therapy is commonly used in clinical practice, which includes cognitive behavioral therapy for insomnia, cognitive behavioral therapy for depression [17], cognitive behavioral therapy for psychosis [18], brief behavioral treatment for insomnia [19], and cognitive behavioral therapy for Parkinson [20]. Among these therapies, cognitive behavioral therapy for insomnia (CBT-I), developed by A.T. Beck in the 1960s, is a branch of CBT and is recommended as first-line treatment for insomnia by the ACP Clinical Guidelines Committee in 2016. e components of CBT-I include cognitive therapy to replace wrong beliefs of sleep; stimulate control to prevent patients from associating with other stimulating activities; sleep restriction to limit time in bed to match perceived sleep duration; sleep hygiene to change habits and physiologic factors and improve sleep; and relaxation to focus on relaxation techniques, such as guided imagery and progressive muscle relaxation (Table 1) [21].Various delivery methods of CBT-I are available, including in-person individual or group therapy, telephone-or Web-based modules, and self-help books. e course of CBT-I varied from 4 sessions to 12 sessions.
We conducted this meta-analysis of randomized controlled trials to assess the clinical effectiveness and safety of CBT-I for insomnia comorbid with depression, compared with no treatment or hypnotics and measured by insomnia outcome measurements, depression outcome measurements, and safety index.

Protocol and Registration.
We made this systematic review and meta-analysis according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA). We registered this review in PROSPERO (CRD42019145065), http://www.crd.york.ac.uk/PROSPERO/.

Exclusion Criteria
(1) Patients could not be diagnosed with insomnia and depression (2) Intervention was not only CBT-I, such as CBT-I plus other therapies (3) Control group was neither no treatment nor hypnotics (4) Outcomes reported incompletely, such as only reported insomnia scales or only reported depression scales (5) Trials did not mention RCT or the word "random" (6) Duplication of the study (7) Study was mechanism, case report, review, or metaanalysis (8) Study language was neither Chinese nor English ey also searched relevant RCTs from existing systematic review and meta-analysis in the reference list. ey downloaded search results for evaluation. ey also contacted authors whose RCTs lacked any relevant information. And they also independently screened the literatures and found out suitable RCTs at the same time, according to inclusion and exclusion criteria. After literature selection, they separately extracted data from these suitable RCTs at the same time. Extracted data included author's name, year of publication, sample size, age, intervention, underlying disease, control group, duration of treatment, time point of assessment, and outcomes. If any disagreement happened, they would resolve it in consultation with more experienced author Yingchun Miao. Yingchun Miao conducted the search, Mei Han evaluated the abstract, and Xun Li evaluated the rest of the paper.

Outcome Measurements.
e outcome measurements of this systematic review and meta-analysis included insomnia outcome measurements, depression outcome measurements, and adverse events. More details of outcome measurements are shown in Table 2.

Risk of Bias.
Guiyu Feng and Le Geng independently evaluated the risk of bias through the Cochrane Handbook for Systematic Reviews of Interventions [22] to evaluate the methodological quality of these included literatures and they performed it via Review Manager 5.3 at the same time. For all evaluation items, the quality of each trial was evaluated using "Yes" (low risk of bias), "No" (high risk of bias), or "Unclear" (unclear risk of bias). Evaluation items included random sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, incomplete outcome data, selective reporting, and other biases. If any disagreement happened, they would resolve it in consultation with more experienced author Yingchun Miao.

Data Synthesis.
We used Cochrane collaboration software RevMan (5.3) to pool outcome data. We calculated the risk ratio (RR) with 95% confidence interval (95% CI) for dichotomous variables and mean difference (MD) with 95% CI for continuous outcomes. Outcomes of insomnia and depression were all continuous variables, and we used mean difference (MD) and its 95% confidence interval (CI) to represent them. Heterogeneity was evaluated by the magnitude of Tau 2 and I 2 statistic. A fixed effect model was performed with minor heterogeneity when the I 2 value was below 50%. A random effect model was used with major heterogeneity when the I 2 value was above 50%. For patient population with underlying diseases, these underlying diseases were different, such as posttraumatic stress disorder (PTSD), ischemic stroke, hypertension, and nonmetastatic cancer. Heterogeneity was major; so, we also performed the random effect model. We would do subgroup analysis and sensitive analysis if the characteristics of data were allowed. If the number of suitable RCTs was more than 10, we would make an inverted funnel plot to assess the impact of publication bias.

Evaluating the Quality of Evidence. Guiyu Feng and Le
Geng independently used the online GRADEpro to assess the quality of evidence (https://gdt.gradepro.org/app/#) at the same time.

Literature Screening and Its Flow Diagram.
We searched 2,101 RCTs according to the search strategy. After duplicated RCTs were deleted, there were 1,641 RCTs left. When we screened the titles and abstracts, 1,387 RCTs which did not meet search criteria were deleted. We then screened the full texts of the remaining 254 RCTs and found 237 RCTs which did not meet search criteria. In the end, 17 RCTs were included in this review. More details of literature screening are shown in Figure 1. Sleep restriction aims to limit time in bed to match perceived sleep duration in order to increase sleep drive and reduce time awake in bed. Time allowed in bed is initially restricted to the average time perceived as sleep per night and then adjusted to ensure that sleep efficiency remains >85%.

Sleep hygiene
Sleep hygiene relates to environmental factors, physiologic factors, and habits that improve sleep, such as regular sleep scheduling, avoiding long daytime naps, and limiting alcohol, caffeine, and nicotine intake especially before bed.

Relaxation
Any relaxation technique that the patient finds effective can be used to limit cognitive arousal and reduce muscular tension to improve sleep. Specific relaxation techniques include meditation, mindfulness, progressive muscle relaxation, guided imagery, and breathing techniques.  Table 3.

Methodological Quality Evaluation.
In order to evaluate the methodological quality of these included literature studies, we used the Cochrane Handbook for Systematic Reviews of Interventions. e overall methodological quality was not good.
For random sequence generation, 15 RCTs [23-35, 37, 39] used the right methods to produce the random sequence, and we assessed them as 'low' risk. One RCT [36] did not mention how to produce the random sequence, except the word 'random,' so we assessed them as 'unclear' risk. 1 RCT [38] used wrong method to produce the random sequence, for it used enrollment order of the facilities, for example, the first enrolled facility was allocated to the CBT-I group while the second to the control group. We assessed it as 'high' risk.
For other biases, we focused on whether the baseline was equal between the intervention group and the control group.
Details of methodological quality evaluation are shown in Figure 2.

Results of Meta-Analysis.
Based on the presence or absence of underlying diseases, patients were divided into 2 groups including patients with underlying diseases and patients without underlying diseases.

Publication Bias.
For the number of included studies in each of these outcomes was less than 10, we could not make an inverted funnel plot to assess the influence of publication bias of included studies.

Evaluating the Quality of Evidence.
e quality of evidence was evaluated via GRADEpro. e quality of evidence in our review varied from moderate to very low. For patients with underlying diseases, the evidence of insomnia outcome measurements was moderate; the evidence of depression outcome measurements varied from low to very low. For patients without underlying diseases, the evidence of insomnia outcome measurements varied from moderate to low; the evidence of depression outcome measurements varied from low to very low. More details are shown in Table 4. e recommendation level assessed by the GRADE system is based on the factors including the advantages of CBT-I, the evidence quality in our review, the preferences of patients, and the cost of CBT-I. e evidence quality varied from moderate to very low in our review; CBT-I is receptible for the patients due to its noninvasive characteristics. And the cost of CBT-I is not so frightfully expensive. So, the recommendation level of the evidence in our review is low.

Summary of Results.
Insomnia scale, depression scale, and adverse event were included in our review to assess the effectiveness and safety of CBT-I. Based on the presence or absence of underlying diseases, we divided patients into 2 groups including patients with underlying diseases and patients without underlying diseases. Our results showed that CBT-I was an effective therapy for insomnia, while CBT-I was not an effective therapy for depression in patients suffering from insomnia and depression. And CBT-I was as effective as hypnotics (zopiclone, estazolam, or benzodiazepine agonist) for insomnia, and both CBT-I and hypnotics were not effective for depression. CBT-I was likely to be a safe therapy due to its noninvasive characteristics. e methodological quality was not good enough. e evidence quality varied from moderate to very low, and the recommendation level based on the evidence was low.

Comparison with the Previous Study.
A number of reviews mentioned the effectiveness of CBT for insomnia or depression, respectively. For example, one review [40] assessed the effectiveness of self-help CBT-I for insomnia by comparing it with waiting list control, routine care, or no treatment, and the result showed that self-help CBT-I was significantly more effective than waiting list control, routine care, or no treatment; another review [41] assessed the     However, few reviews focused on CBT for insomnia comorbidity with depression; therefore, our review assessed the effectiveness of CBT-I for insomnia comorbidity with depression by comparing it with no treatment or hypnotics. And the results of our review showed CBT-I was an effective therapy for patients with insomnia comorbid with depression to some degree. And the clinical effectiveness of CBT-I and hypnotics was familiar with no significant difference between CBT-I and hypnotics.

Strengths and Limitations.
As we mentioned above, a number of reviews focused on the effectiveness of CBT-I for insomnia alone or depression alone. Considering insomnia often co-occurred with depression clinically, there were few guidelines or reviews focused on CBT-I for insomnia comorbidity with depression, so our review assessed the effectiveness of CBT-I for insomnia comorbidity with depression to provide evidence for clinical practice.
Limitations at review level:the quality of evidence varied from moderate to very low; the number of studies in each outcome was less than 10, we could not make publication bias. Limitations at study level: the number of qualified RCTs was insufficient, and the sample size of included studies was IV, random, 95% CI Heterogeneity: tau 2 = 2.15; chi 2 = 2.94, df = 1 (P = 0.09); I 2 = 66% Test for overall effect: Z = 2.29 (P = 0.02)

PSQI
Heterogeneity : tau 2 = 0.33; chi 2 = 4.14, df = 2 (P = 0.13); I 2 = 52%  small. Limitation at outcome level: only one RCT mentioned the item of adverse event and showed that no harm occurred in both CBT-I and no treatment; more evidence is needed to confirm the safety of CBT-I.

Implications for Clinical
Practice. e evidence of our review supported that CBT-I was effective for insomnia in patients suffering from insomnia and depression. And the effectiveness was comparable to hypnotics. CBT-I was likely to be safe due to its noninvasive nature. e findings suggest that CBT-I confers beneficial effects. Because of the low recommendation level of evidence, practitioners could recommend this therapy to patients and finally make a decision based on the evidence, the experience of doctors, and the preferences of patients.

Implications for Future Research.
(1) Future studies should be conducted according to the Consolidated Standards of Reporting Trials (CONSORT) statement, which is essential to control the risk of bias. For example, considering the characteristics of CBT-I, it is hard to blind the doctors and patients, but we could blind outcome assessors. We could also pay more attention to allocation concealment to improve methodological quality of future studies. (2) Researchers should report every detail of studies according to the CONSORT statement, for example, adverse events in CBT-I should be comprehensively reported, although CBT-I appeared to be safe.

Conclusion
CBT-I may be an effective therapy for insomnia, and the effectiveness of CBT-I is comparable to hypnotics, while CBT-I is not an effective therapy for depression, which is the same to hypnotics in patients suffering from insomnia and depression. And CBT-I is likely to be safe. However, the quality and quantity of eligible RCTs are not good enough.
And the evidence quality varied from moderate to very low, and the recommendation level of evidence was low (↑?/2). erefore, more well-designed trials are needed to confirm our findings.