Introduction
Insomnia is characterized by difficulty initiating or maintaining sleep or non-restorative sleep. The disorder often persists for many years (Morin et al. Reference Morin, Bélanger, LeBlanc, Ivers, Savard, Espie, Mérette, Baillargeon and Grégoire2009a ) and is an important public health issue. The prevalence is high, with about a third of the population suffering from insomnia symptoms and about 10% fulfilling the criteria for a sleep disorder (Ohayon, Reference Ohayon2002; Morin et al. 2006b). Insomnia also has a high burden of disease. People with insomnia often report a decline in cognitive abilities and mood swings due to fatigue, which impacts their daily life in various domains (Roth & Ancoli-Israel, Reference Roth and Ancoli-Israel1999; Kyle et al. Reference Kyle, Espie and Morgan2010). Not only is insomnia a significant public health problem in itself, but also many people with insomnia have co-morbid (mental) health problems or will develop co-morbid disorders in the future. Most often reported is the association with depression (Taylor et al. Reference Taylor, Lichstein, Durrence, Reidel and Bush2005; Staner, Reference Staner2010). In addition, research over the past decade provides increasing evidence that insomnia contributes to the risk of developing heart disease (Redline & Foody, Reference Redline and Foody2011) and is associated with increased mortality (Gallicchio & Kaleson, Reference Gallicchio and Kalesan2009). The societal costs due to insomnia are substantial. These costs are caused by increased health-care use, which is about three times higher among poor sleepers than among good sleepers. Most of the societal costs of insomnia (75%) stem from work absenteeism and poor work productivity (Daley et al. Reference Daley, Morin, LeBlanc, Grégoir, Savard and Baillargeon2009b ). In total, poor sleepers cost society about 10 times more than good sleepers (Daley et al. Reference Daley, Morin, LeBlanc, Grégoir and Savard2009a ).
Treatment frequently consists of sleep medication such as benzodiazepines. Several meta-analyses have shown that benzodiazepines are effective in enhancing sleep in the short term (Buscemi et al. Reference Buscemi, Vandermeer, Friesen, Bialy, Tubman, Ospina, Klassen and Witmans2007), but also that there are important side-effects such as drowsiness, dizziness and light headedness. These side-effects increase the risk of (traffic) accidents and, especially in the elderly, the risk of falls. The other treatment option for insomnia is cognitive behavioral therapy (CBT). It has been demonstrated convincingly that CBT is at least as effective as benzodiazepines in the short term, whereas in the longer term CBT is more effective than sleep medication (Smith et al. Reference Smith, Perlis, Park, Smith, Pennington, Giles and Buysse2002; Morin et al. 2009b). However, although CBT is the preferred treatment according to several guidelines (Siebern & Manber, Reference Siebern and Manber2011), it is often unavailable. There is a shortage of CBT therapists, and professionals are often unaware of the treatment facilities that do exist. Moreover, CBT is relatively costly in the short term compared with medication, even though the costs in the longer term (e.g. days away from work) may be reduced by CBT.
A recent development in the management of insomnia is to deliver the treatment over the Internet. During the past decade eHealth has been introduced in mental health care in general. Many Internet-delivered programmes have been developed for different disorders such as depression, alcohol and anxiety disorders (Andrews et al. Reference Andrews, Cuijpers, Craske, McEvoy and Titov2010; Griffiths et al. Reference Griffiths, Farrer and Christensen2010; Riper et al. Reference Riper, Spek, Boon, Conijn, Kramer, Martin-Abello and Smit2011) and meta-analyses have demonstrated that these programmes are effective (Cuijpers et al. Reference Cuijpers, Donker, van Straten, Li and Andersson2010). However, they seem to be much more effective when delivered with some form of guidance and coaching (Spek et al. Reference Spek, Cuijpers, Nyklicek, Riper, Keyzer and Pop2007). Recently, the benefits of coaching in self-help treatments have also been demonstrated for insomnia (Jernelöv et al. Reference Jernelöv, Lekander, Blom, Rydh, Ljótsson, Axelsson and Kaldo2012).
In a previously performed meta-analysis on self-help CBT for insomnia (van Straten & Cuijpers, Reference van Straten and Cuijpers2009), we showed medium effects (e.g. sleep efficiency, Cohen's d = 0.42). However, of the 10 included studies, only one provided the self-help through the Internet (Ström et al. Reference Ström, Pettersson and Andersson2004). Other studies on Internet-delivered CBT (Suzuki et al. Reference Suzuki, Tsuchiya, Hirokawa, Taniguchi, Mitsuhashi and Kawamai2008; Ritterband et al. Reference Ritterband, Thorndike, Gonder-Frederick, Magee, Bailey, Saylor and Morin2009; Vincent & Lewycky, Reference Vincent and Lewycky2009; Espie et al. Reference Espie, Kyle, Williams, Ong, Douglas, Hames and Brown2012; Lancee et al. Reference Lancee, van den Bout, van Straten and Spoormaker2012) have concluded that this treatment has positive effects, although the interventions in these studies were provided without personal guidance or coaching and some studies had small sample sizes.
In the current study we examined the effectiveness of an Internet-delivered CBT guided by a personal coach for adults with insomnia using a randomized controlled trial (RCT).
Method
Design
In our RCT people were randomized to either the intervention (Internet-based CBT) or to a wait-list control group. The wait-list control group received the intervention immediately after the post-test assessment.
Recruitment of patients
In a previous trial, for recruitment purposes, we created a popular scientific website on insomnia (www.insomnie.nl; Lancee et al. Reference Lancee, van den Bout, van Straten and Spoormaker2012). Because of the overwhelming response to this website, we had to create a waiting list. For the current trial we approached those on this waiting list (about 1 year later) by email, in batches of 100 to 200 people to prevent overloading the coaches. The email contained a link to a website with information on this particular trial. On this website, people could register for participation. We sent emails to the first 1500 people on this list as this was the maximum number of patients we could coach.
Inclusion and exclusion criteria
Inclusion criteria were: age ⩾18 years and suffering from insomnia. Insomnia was defined according to DSM-IV criteria as difficulty initiating sleep or difficulty maintaining sleep and based on self-report. To be included people had to be awake for at least 30 min a night, for at least three nights a week, for at least 3 months (APA, 2000). Exclusion criteria were: severe symptoms of anxiety or depression. Anxiety was assessed with the anxiety subscale of the Hospital Anxiety and Depression Scale (HADS; Spinhoven et al. Reference Spinhoven, Ormel, Sloekers, Kempen, Speckens and Van Hemert1997; Bjelland et al. Reference Bjelland, Dahl, Haug and Neckelmann2002; Andersson et al. Reference Andersson, Kaldo-Sandström, Ström and Strömgren2003), which contains seven items. Depression was assessed with the Center for Epidemiological Studies – Depression scale (CES-D; Beekman et al. Reference Beekman, Deeg, Van Limbeek, Braam, De Vries and Van Tilburg1997). People with a HADS score of ⩾10 or a CES-D score of ⩾30 were excluded. Use of sleep medication was allowed.
Procedure
A total of 275 people (18.3%) registered on the trial website (Fig. 1). They were sent an information folder, a consent form, a link to the baseline questionnaire and a 7-day sleep diary. Of these 275 who had registered, 137 (49.8%) returned the informed consent form, the sleep diary and the questionnaire. Of these 137 people, 19 (13.9%) exceeded the cut-off score for depression or anxiety and were excluded. The remaining 118 people (7.9% of the 1500 initially invited) were included in the trial. Half of them (n = 59) were randomized to the guided Internet-based intervention and the other half (n = 59) were randomized to the wait-list control group. Six weeks later they all received another email with a link to the post-test questionnaire and a 7-day sleep diary. Two weeks later the people in the wait-list control group could start the intervention. Eight weeks later we asked the intervention group to again complete a questionnaire and a sleep diary. At this time, the controls were only asked to complete a questionnaire about their experiences with the treatment.
Fig. 1. Flow of participants through the study.
Enrolment took place between May 2010 and September 2010. The study was approved by the Medical Ethical Committee of the Vrije Universiteit Medical Centre, Amsterdam, The Netherlands. The trial was registered on nederlandstrialregister.nl (number NTR2132).
Power and randomization
For pragmatic reasons we were not able to recruit for more than 5 months. Initially, we expected to be able to include only about 50 patients in this period. In other words, we expected to carry out a pilot study and hence no formal power analysis was performed. However, because recruitment was easier than expected, we continued inclusion until we met our full capacity to coach the patients in the intervention.
A randomization schedule was generated by the study coordinator (A.v.S.) by computer. We used blocks of 10 to enhance equal distribution between the groups. The actual randomization was performed by an independent researcher. All patients were informed by email about the randomization outcome.
Primary outcome measure
The patients were asked to complete a sleep diary for 7 days pre- and post-test, and at follow-up for the intervention group. From these diaries we calculated the sleep efficiency (SE), total sleep time (TST), sleep onset latency (SOL) and number of awakenings (NA). SE is calculated by dividing TST by the total time the person spent in bed (×100%). This SE might be considered as the true primary outcome because this measure can be used for patients presenting with different types of sleep problems (Morin, Reference Morin2003). Patients were also asked to rate daily how sound their sleep had been the previous night and how refreshed they felt in the morning. Both questions could be answered on a 10-point scale ranging from 1 (not sound/refreshed at all) to 10 (very sound/refreshed).
Other outcome measures
All patients were asked to complete an online questionnaire pre- and post-test, and at follow-up for the intervention group. With this questionnaire we measured (along with demographics): the duration of sleep problems, overall sleep quality, use of sleep medication, anxiety, depression and quality of life. Overall sleep quality was measured with the Dutch version of the Pittsburgh Sleep Quality Index (PSQI). The questionnaire is well validated in different languages (Buysse et al. Reference Buysse, Reynolds, Monk, Berman and Kupfer1989; Backhaus et al. Reference Backhaus, Junghanns, Broocks, Riemann and Hohagen2002) but Dutch validation studies are lacking. The PSQI is a self-rating questionnaire with 19 questions, and consists of seven subscales (sleep quality, sleep latency, sleep duration, habitual sleep efficiency, sleep disturbance, use of sleep medication and daytime dysfunction). Each subscale is scored on a scale of 0 to 3. The subscale scores can be summed to a total score ranging from 0 (good quality of sleep) to 21 (very poor quality of sleep). As mentioned earlier, the symptoms of anxiety were measured with the HADS and the symptoms of depression with the CES-D. The total score of the seven HADS items range from 0 (no symptoms of anxiety) to 21 (severe symptoms of anxiety). The total score of the 20 CES-D items range from 0 (no symptoms of depression) to 60 (severe symptoms of depression). Quality of life was assessed with one question (on a the visual analogue scale, VAS), which is part of the EuroQoL, in which the patients rate their quality of life on a scale from 0 (poor) to 100 (excellent) (Brooks, Reference Brooks1996). Online administration of questionnaires has been found to generate valid and reliable data with maintained psychometric properties.
After the treatment period all patients were asked to rate the intervention on several aspects. They could give an overall rating score for the treatment itself, and the feedback as provided by the coaches, each on a scale between 1 (very poor) and 10 (excellent). For each of the six lessons we asked whether the information in that particular lesson had been useful (yes/no). We also proposed two statements (‘I gained new insights because of this treatment’ and ‘I'm better able to cope with my sleep problems because of the treatment’) and asked if the patient agreed with those statements (answers on a five-point scale ranging from ‘fully disagree’ to ‘fully agree’). These questions were developed by the research group.
The treatment
The Internet intervention was written (information, examples and assignments) by the first author (A.v.S.). It is based on a collection of other self-help materials for insomnia, textbooks and research literature. The first version of the intervention was discussed with the co-authors and several other sleep experts. The treatment consisted of six weekly lessons and included the different elements that are commonly incorporated in face-to-face CBT for insomnia (Edinger & Wolgemuth, Reference Edinger and Wolgemuth1999; Morin & Espie, Reference Morin and Espie2003; Edinger & Means, Reference Edinger and Means2005; Verbeek & Klip, Reference Verbeek and Klip2005; Espie, Reference Espie2006; Table 1). Every lesson contained information, examples of other people carrying out the treatment, and homework. After finishing the homework, the coach received a notification. Within 3 working days the coach provided online feedback on the homework. Patients could also send separate emails, for example when they had a question about the information provided. At the start of the study, feedback took about 20–30 min per person per lesson. During the study, as the coaches became more experienced, feedback took on average 15 min per person per lesson. The coaching was performed by A.v.S., four master's students in psychology, and one experienced CBT therapist (J.E.) who also trained and supervised the others. The aim of the feedback was to comment on the exercise, clarify information and motivate the patient to persist in carrying out the course and the requested behavioural changes.
Table 1. Overview of the behavioural intervention for insomnia
Analysis
All post-test analyses were performed on the intention-to-treat (ITT) sample. Missing values were imputed using the multiple imputation procedure implemented in SPSS version 20 (SPSS Inc., USA). We created 20 imputation sets. We then used statistical analyses to create pooled estimates of the 20 datasets. Standard deviations for the post-test mean scores were calculated separately because SPSS does not provide them. These were calculated by taking the natural logarithm of the standard deviation of each dataset, and taking the exponent of the summed result divided by the 20 datasets.
Comparisons between baseline values for the intervention and control groups were made with χ 2 tests (for dichotomous variables) or independent t tests (for continuous variables). To test the effectiveness of the intervention we first tested the differences in the post-test scores between the groups with linear regression while controlling for baseline scores and gender. The differences between the two groups were then expressed in effect sizes. We calculated Cohen's d by dividing the difference in post-test scores of the two groups by the pooled standard deviation. Cohen's d can thus be interpreted as the number of standard deviations the intervention group scores better than the control group (Cohen, Reference Cohen1988). A Cohen's d of 0.00–0.32 can be considered as small, 0.33–0.55 as medium, and > 0.56 as large (Lipsey, Reference Lipsey1990; Lipsey & Wilson, Reference Lipsey and Wilson1993).
Next, we wanted to test the clinical relevance of the intervention by comparing the intervention and control groups with regard to the percentage of patients who had (1) improved between baseline and post-test and (2) recovered post-test. However, no consensus consists on the definitions of ‘improved’ or ‘recovered’ for many of the variables studied. Instead of not studying clinical relevance at all, we arbitrarily defined ‘improvement’ and determined post-test thresholds as a proxy for recovery. For the PSQI we defined an improvement as a decrease in score of ⩾3 points. A score > 5 is usually considered an indicator of relevant sleep disturbances (Buysse et al. Reference Buysse, Reynolds, Monk, Berman and Kupfer1989; Backhaus et al. Reference Backhaus, Junghanns, Broocks, Riemann and Hohagen2002). However, only very few people scored below this cut-off. We therefore set the threshold at a score of 8. For SOL we defined improvement as a decrease of ⩾30 min. The post-test threshold for SOL was set at 30 min because there is some consensus that this can be considered ‘normal’ (Lichstein et al. Reference Lichstein, Durrence, Taylor, Bush and Riedel2003). Improvement for TST was defined as sleeping at least 1 h longer and the threshold was set at 6 h. For SE improvement was defined as at least a 10% increase and the threshold was set at 80%. For NA we defined improvement as waking up at least one time less and the threshold was defined as two awakenings. Feeling refreshed and soundness of sleep were both scored on a scale of 1 to 10. Improvement was defined as scoring at least 1 point higher and we set the threshold at 6. Improvement for anxiety was defined as ⩾3 points and for depression and quality of life as ⩾5 points. For the anxiety score on the HADS we used a threshold of ⩽7 (Olsson et al. Reference Olsson, Mykletun and Dahl2005), for the depression score on the CES-D ⩽16 (Beekman et al. Reference Beekman, Deeg, Van Limbeek, Braam, De Vries and Van Tilburg1997) and for quality of life (EuroQoL) ⩾60.
Finally, we tested the robustness of the longer-term effects of the intervention. We did not impute the missing values of the follow-up scores but only used the data as provided by patients. We calculated the effects between post-test and follow-up within the intervention group with Cohen's d.
Results
Baseline characteristics
Most of the patients were female (70.3%), living with a partner (63.6%), born in The Netherlands (84.7%), had a high educational level (58.5%) and a paid job (70.3%; Table 2). There were significantly less females in the intervention group (59.3%) than in the wait-list control group (81.4%, p < 0.01). There were also less people living with a partner in the intervention group (55.9%) than in the control group (71.2%), but this difference only reached borderline significance (p = 0.09).
Table 2. Baseline characteristics of the sample (n = 118)
SOL, Sleep onset latency; TST, total sleep time; SE, sleep efficiency; NA, number of awakenings; s.d., standard deviation.
On average, it took patients almost an hour to fall asleep (mean SOL = 57.1 min), they slept for 5½ h (TST), woke almost twice during the night (NA = 1.9) and slept 67.5% of the time that they were in bed (SE). They rated the soundness of sleep and the feeling of being refreshed as insufficient (5.7 and 5.5 respectively). On average, their sleep problems had existed for 11.8 years (s.d. = 10.2).
In general, it took people in the intervention group longer to fall asleep (SOL 68.7 min) than those in the control group (45.4 min; p < 0.01), mainly because the number of people lying awake for a very long time (⩾ 2 h) was higher in the intervention group (n = 10) than in the control group (n = 2). Furthermore, people in the intervention group woke less often during the night (NA = 1.7) than people in the control group (p = 0.02). There were no significant differences with respect to the use of sleep medication, depression, anxiety or quality of life.
Adherence and satisfaction
Three of the 59 patients (5.1%) in the intervention group did not start the treatment, six (10.2%) completed one or two lessons, seven (11.9%) completed between three and five lessons, and the majority (72.9%; n = 43) completed all six lessons. Most of the patients that dropped out of treatment did not provide any reasons. Some indicated that they were too busy. The treatment was rated as a 7.3 (s.d. = 1.2) on a scale from 1 to 10, and the feedback as a 7.6 (s.d. = 1.2). The third lesson, which was about stimulus control and sleep restriction, was viewed as useful most often (by 79.6% of the patients). All the other lessons were viewed as useful by about 60% of the patients. About two-thirds (61.2%) of the patients agreed with the statement ‘I gained new insights because of this treatment’, and about two-thirds (65.3%) agreed with the statement ‘I'm better able to cope with my sleep problems because of the treatment’.
Post-test effects of the intervention: continuous outcomes
The overall post-test response was 86.4% (n = 102) for the questionnaire, and 71.2% (n = 84) for the sleep diary. The non-response for the sleep diary was significantly higher in the intervention group (n = 22; 37%) than in the control group (n = 12; 20%; p = 0.04). Furthermore, those who returned the sleep diary were more often born in The Netherlands (89% v. 74%, p = 0.03) and they had a shorter SOL at baseline (48.9 min v. 78.0 min, p < 0.01). There were no other significant differences between the responders and the non-responders for other baseline variables or demographics.
The post-test scores of the patients in the intervention group were significantly (p < 0.05) better than those of the control group on all sleep estimates and other health outcomes, after correcting for baseline values and gender, with the exception of SOL (p = 0.14) and NA (p = 0.15; Table 3). The effect sizes for overall sleep quality (the PSQI score), TST, SE, soundness of sleep and quality of life were large (Cohen's d = 1.06, 0.57, 0.95, 0.88 and 0.58 respectively). The effect sizes for NA, feeling refreshed, anxiety and depression were medium (Cohen's d between 0.40 and 0.54). The effect size for SOL was almost absent (d = 0.04). The 95% confidence intervals around the effect sizes are all fairly wide.
Table 3. Post-test effects on sleep and other health outcomes
PSQI, Pittsburgh Sleep Quality Index; SOL, sleep onset latency; TST, total sleep time; SE, sleep efficiency; NA, number of awakenings; s.d., standard deviation; CI, confidence interval.
a Based on imputed means.
b Based on linear regression with baseline values and gender as covariates.
c Post-test comparison of control and intervention groups based on imputed mean scores.
Post-test data on the use of sleep medication were available for 102 (86.4%) of the 118 included patients. There was no statistically significant difference in response between those patients who used sleep medication at baseline (response rate 91.7%) and those that did not (response rate 84.1%; p = 0.27). In the intervention group the use of sleep medication decreased from 28.8% at baseline to 22.4% at post-test (−6.4%). In the control group it increased from 32.2% at baseline to 34.0% at post-test (+1.8%). The post-test difference between the intervention and control group did not reach statistical significance (p = 0.20).
Clinical relevant changes: percentage of patients who improved or scoring below a threshold
We arbitrarily defined improvement, and set thresholds for each of the outcome variables as proxy measurements for recovery. The effects were largest on overall sleep quality and SE: about 60% of the patients in the intervention group improved compared to around 15% in the control group (Table 4). Furthermore, about 50% of the patients in the intervention group scored below the post-test cut-off whereas this was the case for only 9.5% (sleep quality) and 18% (SE) of the control group. The percentage of patients scoring below the cut-off for TST, feeling refreshed and sleeping soundly were also significantly higher in the intervention than in the control group, and they also improved more often. For SOL, NA and anxiety, the differences between the groups were not statistically significant. Patients in the intervention group did improve more often on depression and quality of life than patients in the control group but there was no difference in the percentage of patients scoring below the threshold as this percentage was already high in the control group.
Table 4. Clinically relevant changes: percentage of patients who improved and percentage of patients scoring below predefined thresholds at post-test
PSQI, Pittsburgh Sleep Quality Index; SOL, sleep onset latency; TST, total sleep time; SE, sleep efficiency; HADS, Hospital Anxiety and Depression Scale; CES-D, Center for Epidemiological Studies – Depression scale; NA, number of awakenings; s.d., standard deviation; OR, odds ratio; CI, confidence interval.
a Based on imputed values n = 59.
b Based on logistic regression analyses with gender and baseline values as covariates.
Effects at follow-up
The response at follow-up was 49% (n = 29) for the sleep diary and 73% (n = 43) for the questionnaire. There were no statistically significant baseline differences between the group who did complete a sleep diary at follow-up and those who did not, although people who did not return the diary had borderline significantly higher anxiety and depression scores (p = 0.09 and p = 0.10 respectively).
We compared the observed follow-up scores for patients in the intervention group with post-test scores to obtain some indication of the robustness of effects in the long term (Table 5). None of the improvements or deteriorations were statistically significant and all were very small. Most notable were the continued small improvements in overall sleep quality and quality of life (Cohen's d = 0.25 and 0.21 respectively).
Table 5. Within-group comparison of post-test and follow-up for the intervention group
PSQI, Pittsburgh Sleep Quality Index; SOL, sleep onset latency; TST, total sleep time; SE, sleep efficiency; NA, number of awakenings; s.d., standard deviation; CI, confidence interval.
a Based on imputed means.
b Observed values only.
c Within intervention group comparison of reported post-test and follow-up means.
Discussion
In this RCT we studied the effectiveness of a guided Internet-based CBT for adults with insomnia. First, we showed that adherence was good (72.9% completed the intervention) and that the patients were satisfied with the information they received, the assignments, the feedback from the coaches and the effects of the intervention. Second, we did not find any effect of the intervention on SOL and NA, possibly because of baseline differences and because SOL had some outliers. Third, the treatment had significant effects (p < 0.05) on all the remaining sleep estimates (sleep quality, TST, SE, sleeping soundly and feeling refreshed in the morning) and on the secondary outcomes (symptoms of anxiety and depression, quality of life). The effect sizes were medium to large (Cohen's d between 0.40 and 1.06). Patients in the intervention group more often improved and reached recovery than those in the control group. The use of medication decreased in the intervention group by 6.8% and increased in the control group by 1.8%. However, these analyses were performed on the sample of completers and were not statistically significant. Finally, the completers-only sample did not show any significant improvement or deterioration in the longer term (3 months).
There were some baseline differences between the intervention and baseline groups (gender, SOL, NA). We therefore used gender as a covariate in all the statistical tests on post-test differences, and we also included baseline values as covariates. The post-test response was high, with 88.1% of the patients providing some or all of the data. Although there were some baseline differences between post-test responders and non-responders, the risk of bias is probably low because the non-response percentage was small and these data were estimated using multiple imputation. However, the follow-up data should be interpreted with caution. As the non-response percentage for the sleep diary was 51%, we decided not to impute the data but to show the results for the responders only. Hence these data may be biased, and suggest that more research on the longer-term effects of Internet treatments for insomnia is necessary.
We recruited our patients through a waitlist, which comprised people who had indicated 1 year earlier that they were interested in participating in an insomnia study. Those people that responded to our invitation were still, or again, suffering from insomnia. Thus, it is thus likely that our sample was skewed towards higher insomnia severity. Baseline sleep estimates indeed indicate severe problems: SOL of almost 1 h, TST of 5.5 h and SE of 68%, which on average had lasted for 12 years. Even though this might indicate that our group is not representative for all insomnia patients, it might consist of the patients most in need for treatment.
The effects of our intervention are promising. Two previous meta-analyses on self-help for insomnia showed, for example, an effect size of 0.40 (Cheng & Dizon, Reference Cheng and Dizon2012) and 0.42 (van Straten & Cuijpers, Reference van Straten and Cuijpers2009) for SE whereas in the current study the SE effect size was much higher (d = 0.95). One reason for our positive results might be that the patients in the intervention received regular weekly feedback from their personal coach. This might also have been responsible for our high adherence rates and overall satisfaction with the intervention. In general, higher effects are demonstrated for guided web-based interventions than for unguided ones (Spek et al. Reference Spek, Cuijpers, Nyklicek, Riper, Keyzer and Pop2007; Richards & Richardson, Reference Richards and Richardson2012). However, it has been argued that unguided interventions might in future become as beneficial as guided interventions once the websites are more interactive and technically better, for example including automated personalized feedback based on text or answers on quizzes and questionnaires. Some studies on unguided Internet-delivered insomnia treatments also show promising results (e.g. Ritterband et al. Reference Ritterband, Thorndike, Gonder-Frederick, Magee, Bailey, Saylor and Morin2009; Espie et al. 2012) but, to date, only one study has compared guided with unguided self-help for insomnia (Jernelöv et al. Reference Jernelöv, Lekander, Blom, Rydh, Ljótsson, Axelsson and Kaldo2012). That study supports the notion that guidance increase effectiveness. We recommend replication of that study using Internet-guided treatment and also examining the cost-effectiveness of the two approaches.
The effects of face-to-face treatments for insomnia are well studied and their results are summarized in reviews (e.g. Morin et al. 2006a). Although no formal meta-analysis has been performed and no overall estimate for face-to-face treatments is available, the effects seem to be of the same order of magnitude as those of our study. Unfortunately, there are very few studies that directly compare face-to-face treatments with self-help or Internet treatments. In our previous meta-analysis on self-help, we demonstrated that those few studies that exist do not demonstrate a clear difference in effect (van Straten & Cuijpers, Reference van Straten and Cuijpers2009). The comparability of effects between face-to-face treatments and self-help (Internet) treatments has been demonstrated for anxiety and depression (Cuijpers et al. Reference Cuijpers, Donker, van Straten, Li and Andersson2010). We need further studies examining accessibility, effects and costs, to demonstrate which intervention should be used when and for whom.
In our opinion people with co-morbidity should be included in insomnia treatment trials because co-morbidity tends to be the rule rather than the exception. In particular, co-morbidity with mental disorders is very common. People with insomnia are about 10 times more likely to have depression and 17 times more likely to have an anxiety disorder than people without insomnia (Taylor et al. Reference Taylor, Lichstein, Durrence, Reidel and Bush2005). In our study we excluded people with severe anxiety or depressive disorders because the most effective treatment strategy for people with co-morbid insomnia and more severe mental health problems is not known. This is an important topic that requires further investigation. We did include people who reported moderate symptoms of anxiety and depression. These symptoms improved during the intervention period with medium effect sizes (Cohen's d = 0.53 and 0.44). This might indicate that insomnia is one of the causes of mental health problems or that there are other underlying mechanisms that lead to disruption of both mood and sleep (Turek, Reference Turek2005; Fairholme et al. Reference Fairholme, Carl, Farchione and Schonwetter2012). This significant finding stresses the importance of insomnia treatment, as it might be useful to reduce moderate symptoms of depression or anxiety.
There are several limitations to this study. First, the diagnosis of insomnia was based purely on self-report and was not confirmed by a clinician. This means that it is possible that some people in our sample did not suffer from full-blown insomnia or suffered other sleep disorders (e.g. sleep apnoea). In future we would prefer this intervention to be delivered through general practitioners (GPs). This means that the GPs might screen out mild cases and those with other serious sleep or medical disorders. However, we would like to stress that our method of recruitment did not result in a sample of only mild cases. The fact that 72.9% of the patients completed the intervention (and put in considerable effort in completing the exercises) seems to indicate that those people were indeed in need of help. A second limitation is that the sleep estimates were based on sleep diaries and not on more objective measures such as polysomnography or actigraphy. The use of both subjective and objective measures has been recommended because people with insomnia often over- or underestimate their actual sleep time (Buysse et al. Reference Buysse, Ancoli-Israel, Edinger, Lichstein and Morin2006; Van den Berg et al. Reference Van den Berg, Van Rooij, Vos, Tulen, Hofman, Miedema, Knuistingh Neven and Tiemeier2008). However, using polysomnography is costly and imposes a burden on the patients. Therefore, sleep diaries are currently the most widely used outcome measure in insomnia treatment studies (Morin, Reference Morin2003). Sleep diaries are also generally well accepted because it is the subjective complaint that prompts patients to seek treatment. The third limitation is that we did not measure daytime consequences of insomnia. As it is already an effort for patients to keep a sleep diary, we wanted to keep the number of remaining questions as low as possible. However, now that this intervention has proved to be effective with regard to sleep estimates, a next step would be to investigate the consequences of these improvements for daytime functioning. This is ultimately the most important outcome for patients but is also essential in demonstrating possible cost-effectiveness. Almost two-thirds of our sample had a paid job, and loss of work productivity is one of the most common consequences of insomnia but also the most costly (Dailey et al. Reference Daley, Morin, LeBlanc, Grégoir and Savard2009a ). The self-help study of Jernelöv et al. (Reference Jernelöv, Lekander, Blom, Rydh, Ljótsson, Axelsson and Kaldo2012) is one of the few to demonstrate positive effects on daytime functioning after self-help treatment.
In summary, this study adds to the growing body of literature that indicates that guided CBT for insomnia can be delivered through the Internet. We suggest that it is time for large-scale implementation projects.
Acknowledgements
We are grateful for the financial support we received from Fund NutsOhra (0804–46) to develop the intervention and to carry out this study.
Declaration of interest
None.
