Internet-Delivered Cognitive-Behavioral Therapy for Insomnia in Breast Cancer Survivors: A Randomized Controlled Trial

The intervention was a fully automated, internet-delivered cognitive-behavioral therapy for insomnia (iCBT-I) program called "Sleep Healthy Using the Internet" (SHUTi). The program consisted of six sequential "cores" (sessions) delivered over approximately nine weeks. Each core took about 45–60 minutes to complete and included:

• Core 1: Psychoeducation about sleep, the two-process model of sleep regulation (homeostatic drive vs. circadian alerting signal), and the "3P" model of insomnia (predisposing, precipitating, and perpetuating factors).
• Core 2: Sleep restriction therapy — participants were instructed to limit time in bed to match their average total sleep time, then gradually increase as sleep efficiency improved.
• Core 3: Stimulus control — strengthening the association between bed and sleep by going to bed only when sleepy, getting out of bed if awake for more than 15–20 minutes, and using the bed only for sleep and sex.
• Core 4: Cognitive restructuring — identifying and challenging unhelpful beliefs about sleep (e.g., "I'll never function tomorrow if I don't sleep eight hours").
• Core 5: Sleep hygiene and relapse prevention — reviewing caffeine, alcohol, exercise timing, and planning for future sleep disruptions.
• Core 6: Summary and maintenance planning.

The comparator was a waitlist control group — participants received no active treatment during the study period but were offered the iCBT-I program after the follow-up period ended. This is a weaker control than an active placebo or attention-control condition, because waitlist participants know they are not receiving treatment, which can inflate the apparent effect of the intervention.

The primary outcome was insomnia severity measured by the Insomnia Severity Index (ISI, 0–28 scale, higher = worse). Secondary outcomes included sleep quality (Pittsburgh Sleep Quality Index, PSQI), fatigue (Multidimensional Fatigue Inventory, MFI-20), and sleep diary parameters (sleep onset latency, wake after sleep onset, total sleep time, sleep efficiency).

• Sample size: 255 women (133 allocated to iCBT-I, 122 to waitlist control)
• Population: Danish breast cancer survivors who had completed primary cancer treatment (surgery, chemotherapy, and/or radiation) at least three months prior to enrollment
• Inclusion criteria: Clinically significant sleep disturbance defined as an Insomnia Severity Index score ≥ 12; access to a computer with internet; ability to read and understand Danish
• Exclusion criteria: Current or previous diagnosis of other sleep disorders (e.g., sleep apnea, restless legs syndrome); current participation in another insomnia treatment; severe psychiatric comorbidity (e.g., bipolar disorder, psychosis); current shift work; pregnancy
• Setting: National sample recruited through the Danish Breast Cancer Cooperative Group registry and advertisements in patient organizations
• Demographics: Mean age 54.1 years (SD = 9.8, range 28–78); mean time since diagnosis 4.1 years (SD = 3.5); 68% had received chemotherapy, 78% radiation, 72% hormone therapy; 41% were currently taking antidepressants or anxiolytics

• Insomnia Severity Index (ISI): 7-item self-report questionnaire, scored 0–28. Scores ≥ 15 indicate clinical insomnia; scores 8–14 indicate subthreshold insomnia. The minimal clinically important difference is approximately 6 points.
• Pittsburgh Sleep Quality Index (PSQI): 19-item self-report, scored 0–21. Scores > 5 indicate poor sleep quality. Measures seven components: subjective sleep quality, sleep latency, sleep duration, habitual sleep efficiency, sleep disturbances, use of sleep medication, and daytime dysfunction.
• Multidimensional Fatigue Inventory (MFI-20): 20-item self-report measuring five dimensions of fatigue: general fatigue, physical fatigue, reduced activity, reduced motivation, and mental fatigue. Each subscale scored 4–20, higher = more fatigue.
• Sleep diary: Participants completed online sleep diaries for two-week periods at baseline and post-intervention. Diaries captured: bedtime, wake time, rise time, sleep onset latency (SOL), number and duration of night wakings (wake after sleep onset, WASO), total sleep time (TST), and sleep efficiency (SE = TST / time in bed × 100%).
• Actigraphy: A subset of participants (n = 60) wore wrist actigraphs (MotionWatch 8) for one week at baseline and post-intervention to objectively measure sleep parameters. Actigraphy uses movement sensors to estimate sleep/wake patterns.

Design: Two-arm, parallel-group, randomized controlled trial (RCT) with a 55:45 allocation ratio favoring the intervention group. Assessments occurred at baseline, post-intervention (9 weeks), and follow-up (15 weeks from baseline, i.e., 6 weeks post-intervention).

Randomization: Participants were randomly allocated using a computer-generated random sequence with a 1:1 ratio (though the final allocation was 55:45 due to a randomization error — the authors note this was a technical issue, not a bias). Randomization was stratified by age (≤50 vs. >50) and use of sleep medication (yes/no). The allocation sequence was concealed from study staff until after enrollment.

Blinding: This was an unblinded trial. Participants knew whether they were in the active treatment or waitlist group. Outcome assessors were not blinded because all outcomes were self-reported online. The lack of blinding is a significant limitation — participants who know they are receiving an active treatment may report better outcomes due to expectation effects (placebo), while waitlist participants may report worse outcomes due to disappointment or demoralization (nocebo effect).

Duration: The intervention period was 9 weeks (participants were asked to complete one core per week, but could progress at their own pace). The total study duration from baseline to final follow-up was 15 weeks.

Statistical approach: Primary analyses used intention-to-treat (ITT) — all randomized participants were included regardless of how much of the intervention they completed. Mixed linear models were used to test time × group interactions, which accounts for missing data under the assumption that data are missing at random. All tests were two-sided. To control for multiple outcomes, the authors used a Benjamini-Hochberg false discovery rate correction. Effect sizes were reported as Cohen's d (0.2 = small, 0.5 = medium, 0.8 = large).

What this design can and cannot prove:

• Can prove: That iCBT-I causes improvements in self-reported sleep and fatigue outcomes compared to no treatment in breast cancer survivors, over a 15-week period. The RCT design with random allocation controls for confounding variables (age, cancer treatment history, baseline sleep severity) that could otherwise explain differences between groups.
• Cannot prove: That iCBT-I is superior to an active placebo or to in-person CBT-I. The waitlist control cannot distinguish specific treatment effects from non-specific effects (attention, expectation, therapeutic alliance). The design also cannot prove long-term durability beyond 15 weeks, cannot prove effectiveness in men or in non-breast cancer populations, and cannot prove that objective sleep changes (measured by actigraphy) match subjective improvements.

Major methodological weaknesses:

1. No active control condition — waitlist designs inflate effect sizes because of negative expectations in the control group
2. No blinding — all outcomes are self-reported, and participants knew their group assignment
3. Actigraphy data only available for a subset (n = 60), limiting objective validation
4. High attrition — 18% of the iCBT-I group and 11% of the waitlist group dropped out by post-intervention; by follow-up, attrition was 25% and 16%, respectively
5. Short follow-up — 15 weeks is insufficient to assess whether improvements are maintained long-term
6. Single-country sample — all participants were Danish, limiting generalizability to other healthcare systems and cultures

Primary outcome — Insomnia Severity Index (ISI):

• Baseline mean: iCBT-I = 18.8 (SD = 3.9), waitlist = 18.4 (SD = 3.8) — both in the clinical insomnia range (≥15)
• Post-intervention mean: iCBT-I = 9.9 (SD = 5.2), waitlist = 16.1 (SD = 5.2)
• Time × group interaction: P < .001
• Effect size: d = 1.17 (95% CI = 0.87 to 1.47) — a large effect
• At follow-up (15 weeks): iCBT-I = 9.5 (SD = 5.3), waitlist = 15.7 (SD = 5.6); d = 1.10 (95% CI = 0.80 to 1.40)

Secondary outcomes — Sleep diary parameters (post-intervention):

• Sleep onset latency (SOL): iCBT-I decreased from 41.5 to 24.1 minutes; waitlist decreased from 39.8 to 36.4 minutes. Between-group difference: ~12 minutes. Effect size: d = 0.55 (95% CI = 0.27 to 0.83), P = .001
• Wake after sleep onset (WASO): iCBT-I decreased from 62.3 to 44.8 minutes; waitlist decreased from 58.9 to 54.2 minutes. Between-group difference: ~9 minutes. Effect size: d = 0.33 (95% CI = 0.06 to 0.61), P = .02 — a small effect
• Sleep efficiency (SE): iCBT-I increased from 72.1% to 82.5%; waitlist increased from 73.4% to 76.2%. Between-group difference: ~6 percentage points. Effect size: d = 0.68 (95% CI = 0.39 to 0.97), P < .001
• Total sleep time (TST): iCBT-I increased from 6.1 to 6.5 hours; waitlist increased from 6.2 to 6.3 hours. Between-group difference: ~12 minutes. Effect size: d = 0.27 (95% CI = -0.01 to 0.55), P = .06 — not statistically significant

Secondary outcomes — Sleep quality (PSQI):

• Post-intervention: iCBT-I = 8.1 (SD = 3.6), waitlist = 11.7 (SD = 3.8); d = 0.97 (95% CI = 0.68 to 1.26), P < .001
• Follow-up: d = 0.86 (95% CI = 0.56 to 1.15)

Secondary outcomes — Fatigue (MFI-20 general fatigue subscale):

• Post-intervention: iCBT-I = 13.1 (SD = 4.2), waitlist = 15.4 (SD = 4.0); d = 0.56 (95% CI = 0.28 to 0.84), P < .001
• Follow-up: d = 0.66 (95% CI = 0.37 to 0.95)

Objective sleep (actigraphy, n = 60):

• No statistically significant differences between groups for any actigraphy-measured parameter (SOL, WASO, TST, SE). This is important — subjective improvements were not confirmed by objective measurement in the subset that wore actigraphs.

Clinical significance:

• At post-intervention, 54% of the iCBT-I group no longer met criteria for clinical insomnia (ISI < 15), compared to 24% of the waitlist group
• At follow-up, 56% of the iCBT-I group were in remission vs. 27% of the waitlist group

To translate these numbers into plain English:

• Insomnia severity dropped by ~9 points on the 28-point ISI scale. This is well above the minimal clinically important difference of 6 points. The average participant went from "moderate clinical insomnia" to "subthreshold insomnia" — meaning their sleep problems were still present but no longer severe enough to warrant a clinical diagnosis.
• Falling asleep got ~17 minutes faster on average (from 42 minutes to 24 minutes). This is roughly the time it takes to read a short news article — a meaningful reduction in bedtime frustration.
• Night waking time decreased by ~18 minutes (from 62 to 45 minutes). This is about the length of a TV sitcom episode — less time lying awake staring at the ceiling.
• Sleep efficiency improved from 72% to 83% — meaning participants went from spending more than a quarter of their time in bed awake to spending only about one-sixth of their time awake. The clinical target for good sleep is typically ≥85%.
• Total sleep time increased by only ~24 minutes (from 6.1 to 6.5 hours), and this was not statistically significant. This is important: CBT-I often does not increase total sleep time dramatically; it improves the quality and consistency of sleep rather than the quantity.
• Fatigue decreased by ~2.3 points on the 4–20 MFI general fatigue scale — a moderate effect. This is roughly equivalent to the difference between feeling "a bit tired" and "moderately tired" on a typical day.

Acknowledged by authors:

• Waitlist control design cannot control for non-specific treatment effects (attention, expectation, therapeutic alliance)
• No blinding of participants or outcome assessors
• High attrition, particularly in the iCBT-I group (18% at post-intervention, 25% at follow-up)
• Short follow-up period (15 weeks)
• Actigraphy data only available for a subset, limiting objective validation
• All participants were Danish, limiting generalizability to other populations

Additional critical observations:

• No active comparator — the study cannot tell us whether iCBT-I is better than sleep hygiene education, relaxation training, or even a simple sleep tracking app. The effect sizes are likely inflated compared to what would be seen against an active control.
• Self-report bias — all primary and secondary outcomes were self-reported. Participants who invested time in a 6-session program may be motivated to report improvement (demand characteristics). The actigraphy data showing no objective improvement raises questions about whether the subjective improvements reflect actual sleep changes or changed perceptions of sleep.
• Selection bias — participants were volunteers who had internet access and were motivated to complete an online program. This may not represent the average breast cancer survivor, particularly older women or those with limited digital literacy.
• Co-interventions — 41% of participants were taking antidepressants or anxiolytics. The study did not control for changes in medication use during the trial, which could confound results.
• No assessment of sleep apnea — participants with diagnosed sleep apnea were excluded, but no screening was done for undiagnosed sleep apnea. Given that breast cancer survivors are often older and may have gained weight during treatment, undiagnosed sleep apnea could be common.
• Single-sex sample — results may not generalize to male cancer survivors or to non-cancer populations with insomnia.

For someone running their own n=1 experiment to improve sleep:

What to test

• The intervention: A structured, multi-component CBT-I program delivered online or via a workbook