Insomnia as a mediating therapeutic target for depressive symptoms: A sub‐analysis of participant data from two large randomized controlled trials of a digital sleep intervention

Summary Insomnia predicts the onset of depression, commonly co‐presents with depression and often persists following depression remission. However, these conditions can be challenging to treat concurrently using depression‐specific therapies. Cognitive behavioural therapy for insomnia may be an appropriate treatment to improve both insomnia and depressive symptoms. We examined the effects of a fully‐automated digital cognitive behavioural therapy intervention for insomnia (Sleepio) on insomnia and depressive symptoms, and the mediating role of sleep improvement on depressive symptoms in participants from two randomized controlled trials of digital cognitive behavioural therapy for insomnia. We also explored potential moderators of intervention effects. All participants met criteria for probable insomnia disorder and had clinically significant depressive symptomatology (PHQ‐9 ≥ 10; n = 3,352). Individuals allocated to treatment in both trials were provided access to digital cognitive behavioural therapy. Digital cognitive behavioural therapy significantly improved insomnia (p < .001; g = 0.76) and depressive symptoms (p < .001; g = 0.48) at post‐intervention (weeks 8–10), and increased the odds (OR = 2.9; 95% CI = 2.34, 3.65) of clinically significant improvement in depressive symptoms (PHQ‐9 < 10). Improvements in insomnia symptoms at mid‐intervention mediated 87% of the effects on depressive symptoms at post‐intervention. No variables moderated effectiveness outcomes, suggesting generalizability of these findings. Our results suggest that effects of digital cognitive behavioural therapy for insomnia extend to depressive symptoms in those with clinically significant depressive symptomatology. Insomnia may, therefore, be an important therapeutic target to assist management of depressive symptoms.


Summary
Insomnia predicts the onset of depression, commonly co-presents with depression and often persists following depression remission. However, these conditions can be challenging to treat concurrently using depression-specific therapies. Cognitive behavioural therapy for insomnia may be an appropriate treatment to improve both insomnia and depressive symptoms. We examined the effects of a fully-automated digital cognitive behavioural therapy intervention for insomnia (Sleepio) on insomnia and depressive symptoms, and the mediating role of sleep improvement on depressive symptoms in participants from two randomized controlled trials of digital cognitive behavioural therapy for insomnia. We also explored potential moderators of intervention effects. All participants met criteria for probable insomnia disorder and had clinically significant depressive symptomatology (PHQ-9 ≥ 10; n = 3,352). Individuals allocated to treatment in both trials were provided access to digital cognitive behavioural therapy. Digital cognitive behavioural therapy significantly improved insomnia (p < .001; g = 0.76) and depressive symptoms (p < .001; g = 0.48) at post-intervention (weeks 8-10), and increased the odds (OR = 2.9; 95% CI = 2.34, 3.65) of clinically significant improvement in depressive symptoms (PHQ-9 < 10). Improvements in insomnia symptoms at mid-intervention mediated 87% of the effects on depressive symptoms at post-intervention. No variables moderated effectiveness outcomes, suggesting generalizability of these findings. Our results suggest that effects of digital cognitive behavioural therapy for insomnia extend to depressive symptoms in those with clinically significant depressive symptomatology. Insomnia may, therefore, be an important therapeutic target to assist management of depressive symptoms.
Despite effective pharmacological and psychological treatments, depression remains the leading cause of disability worldwide (Friedrich, 2017). Insomnia is a promising therapeutic target in depressive disorders because sleep disturbance is a typically co-morbid characteristic; and because insomnia symptoms often persist with otherwise effective treatment of depression (Nutt, Wilson, & Paterson, 2008).
Moreover, meta-analytic evaluation suggests an aetiological link; individuals with pre-existing chronic insomnia being twice as likely to develop first episode depression (Baglioni et al., 2011).
Cognitive behavioural therapy (CBT) for insomnia is an effective intervention for insomnia and co-morbid depressive symptoms (Gebara et al., 2018). Studies of in-person CBT have shown beneficial effects when delivered standalone or adjunctive to depression treatments (Gebara et al., 2018); however, systemic barriers prevent widespread access to in-person CBT (Koffel, Bramoweth, & Ulmer, 2018).
The evolution of digital CBT formats (web/mobile) address accessibility challenges around managing insomnia .
Nevertheless, the effects of fully-automated digital CBT on depressive symptoms in participants with clinically significant depression symp- tomatology have yet to be examined. Although examined in trials of therapist-delivered CBT (Manber et al., 2016), the mediating role of insomnia on depressive symptoms has also not been investigated using digital CBT. Data from trials of fully-automated digital CBT provide an opportunity to test this hypothesized relationship between sleep and subsequent depressive symptom improvement in large sample sizes.
In this short report, we examined the effects of digital CBT for insomnia compared with control on insomnia and depressive symptoms, and the mediating role of insomnia using a sub-sample of participants with insomnia and clinically significant depression symptomatology from two large effectiveness trials. These trials investigated the mediating role of insomnia on daytime functioning (Espie et al., 2019) and mental health following digital CBT (Freeman et al., 2017) in individuals with insomnia. We also explored potential moderators of intervention effects.

| ME THODS
In both randomized controlled trials (RCTs), participants scoring < 16 on the Sleep Condition Indicator (SCI-8), suggestive of probable DSM-5 insomnia disorder, had been randomized to either Sleepio, a homogeneous fully-automated digital CBT intervention for insomnia, or to a control condition (sleep hygiene education [Espie et al., 2019] or usual care [Freeman et al., 2017] (Espie et al., 2019;Freeman et al., 2017). The primary trials obtained informed consent from all participants. For this sub-analysis we selected all those participants who, in addition to probable insomnia disorder, had a PHQ-9 score ≥ 10 at baseline; this cut-off being suggestive of clinically significant depressive symptoms (National Collaborating Centre for Mental Health, 2018). We also examined the likelihood participants experienced clinically significant improvement (PHQ-9 < 10), and reliable change (PHQ-9 < 10 with change scores ≥ 6) at post-intervention and follow-up (National Collaborating Centre for Mental Health, 2018). Baseline measures are presented using summary statistics, with hypothesis testing to check for balance as the original trials were not stratified for the selection criteria of depression caseness as in this study. Statistical analyses were performed using generalized linear models with maximum likelihood estimation with the following fixed effects: treatment, trial, baseline outcome, and an interaction between baseline and trial to allow for a different baseline effect in each trial. For continuous outcomes this was performed using linear models, for remission analyses (PHQ-9 < 10) and reliable remission (PHQ-9 < 10 plus a minimum reduction of ≥ 6) using logistic regression models and mediation analysis was performed using parametric structural equation models (StataCorp version 16). See Supporting information for further details on mediation analysis. Moderation was assessed by adding interactions between moderator and treatment to the analysis models. All analyses were based on the intention-to-treat principle. First, and consistent with published trial results, digital CBT led to significant improvements in insomnia symptoms (SCI-8) at post-intervention (8-10 weeks; adjusted mean difference 5.19, 95% CI 4.63−5.75, g = 0.76) and follow-up (22-24 weeks; adjusted mean difference 5.15, 95% CI 4.47−5.83, g = 0.69) compared with control. Second, digital CBT also significantly reduced depressive symptoms (PHQ-9) at post-intervention (8-10 weeks; adjusted mean difference −3.03, 95% CI −3.56 to −2.50, g = 0.48) and follow-up (22-24 weeks; adjusted mean difference −2.74, 95% CI −3.34 to −2.15, g = 0.42) compared with control ( Figure 1; Table 1). Removing the sleep item from the PHQ-9 did not alter the effects of digital CBT on depressive symptoms.
Importantly, in relation to the hypothesis of insomnia improvement as an explanatory factor, mediation analyses demonstrated that improvements in sleep (on the SCI-8) at mid-intervention (weeks 3-4) mediated 87% of the intervention effect on depressive symptoms at post-intervention (weeks 8-10; Table 2). With the sleep item removed from the PHQ-9, improvements in sleep (SCI-8) at mid-intervention mediated 91% of the intervention effect on depressive symptoms at post-intervention. We also found that intervention effects were not moderated by age, gender, or by baseline insomnia or depressive symptoms.

| D ISCUSS I ON
Cognitive behavioural therapy is already the guideline treatment for insomnia disorder (Riemann et al., 2017). Our findings indicate that the effects of digital CBT for insomnia are not limited to sleep but extend to improving depressive symptoms, in accordance with other literature (Gebara et al., 2018). Indeed, these results suggest that intervention upon sleep may increase the odds of depression F I G U R E 1 Mean depressive symptoms (a) and insomnia symptoms (b) over time. Note: (a) Average depressive symptoms measured by the 9-item Patient Health Questionnaire (PHQ-9). Outcomes were assessed at baseline, post-intervention (weeks 8 and 10) and follow-up (weeks 22 and 24). Error bars represent standard deviations. (b) Average insomnia symptoms were measured by the 8-item Sleep Condition Indicator (SCI-8). Higher scores on the SCI-8 indicate better sleep improvement (PHQ-9 < 10) by a factor of 2.4−3 times. This may be a causal relationship, because reduction in insomnia symptoms at midintervention (3-4 weeks) mediated the effect of the intervention on depressive symptoms at post-intervention (8-10 weeks). The finding that intervention effects were not moderated by demographics or by more severe baseline symptoms of insomnia or depression supports the generalizability of these data. Putative mechanisms proposed to account for the antidepressant effects of CBT for insomnia include sleep restriction therapy and stimulus control increasing slow-wave sleep, which is known to be disrupted in both depression and insomnia (Krystal & Edinger, 2010;Staner, 2010), thereby increasing homeostatic sleep drive. Additionally, establishment of a consistent sleep−wake schedule and increased time awake during the day may contribute to behavioural activation (Maroti, Folkeson, Jansson-Fröjmark, & Linton, 2011), and cognitive therapy, when applied to insomnia, may result in generalized benefits to mood through a marked shift in perspective when a fundamental dilemma becomes resolved, consistent with cognitive models of depression (Beck, 2002).
Our results should be interpreted with some caution given that they are derived from only two effectiveness RCTs, and these were recruited samples. However, the sample size was large enough to reliably test these relationships, and the homogeneous nature of the intervention usefully reduces the variance typically associated with combining data across diverse CBT approaches to intervention.
Further work is warranted, especially in large clinical samples, to consider sleep as a therapeutic target to improve depressive symptoms, especially in patients with both insomnia disorder and major depressive disorder. Finally, given that population-level access to therapist-delivered CBT is limited due to lack of therapists and other barriers (Koffel et al., 2018), digital CBT may offer an accessible firststep intervention to improve insomnia with attendant clinical benefits to depressive symptoms. Post-intervention SCI-8 mid-intervention