Learning: CBT-I

The American College of Physicians recommends that all adults with chronic insomnia disorder receive CBT-I as the initial treatment, ahead of medication. That is a strong recommendation on moderate-quality evidence. Adding a drug is a separate, weak recommendation resting on low-quality evidence, and is framed as a shared decision after CBT-I alone has been tried. Measured in: Adults with chronic insomnia disorder, across the trials underlying the guideline's systematic review. A strong recommendation on moderate-quality evidence means the direction is settled and the size of the benefit is not. The guideline is from 2016 and predates the component-level and single-component trials that followed it. The guideline’s trigger for considering medication is that CBT-I has been UNSUCCESSFUL, not merely that it has been tried.

The American Academy of Sleep Medicine makes a strong recommendation for multicomponent CBT-I. Stimulus control, sleep restriction therapy and relaxation therapy each carry a conditional recommendation as single-component treatments. Sleep hygiene carries a conditional recommendation against being used as a single-component therapy, though it may sit inside a package. Measured in: Adults with chronic insomnia disorder; a GRADE-assessed systematic review underlying the guideline. Only the multicomponent recommendation is strong. Every single-component recommendation, including sleep restriction on its own, is conditional, which the guideline defines as a suggestion to be weighed against patient values and circumstances, not applied by default.

Across 20 randomized trials against inactive controls (1,162 adults, 64% women, mean age 56), time to fall asleep fell 19.0 minutes (95% CI 14.1 to 23.9), time awake in the night fell 26.0 minutes (15.5 to 36.5), and the share of time in bed actually spent asleep rose 9.9 percentage points (8.1 to 11.7), sustained at follow-up with no adverse outcomes reported. The controls did nothing, so part of the gap is doing something structured versus nothing, and the sleep figures come from self-reported diaries in unblinded trials, which inflates the subjective gains for a treatment this effortful.

In a component network meta-analysis of 241 trials (31,452 people, mean age 45.4, 67% women), four parts were tied to better remission: cognitive restructuring (incremental OR 1.68, 95% CI 1.28 to 2.20), third-wave acceptance and mindfulness work (1.49, 1.10 to 2.03), sleep restriction (1.49, 1.04 to 2.13) and stimulus control (1.43, 1.00 to 2.05). In-person therapist-led delivery added the largest single increment (1.83, 1.19 to 2.81). Because it infers each part from packages that combined them differently, not testing them head to head, and the sleep-restriction and stimulus-control intervals both run close to 1, the ranking is modeled, not proven.

Four brief nurse-delivered sleep-restriction sessions in general practice beat sleep-hygiene guidance at six months: insomnia severity 10.9 versus 13.9, an adjusted difference of 3.05 points (Cohen's d 0.74), at a cost of £2,076 per quality-adjusted life year and a 95.3% chance of being cost-effective at the usual threshold. Eight people in each arm had a serious adverse event, none treatment-related. The trial was open-label, so people knew their arm, and the nurse contact is more support than a written protocol alone gives.

Pooling 30 randomized trials against inactive controls, insomnia severity stayed better than control at 3 months (Hedges g 0.64), 6 months (0.40) and 12 months (0.25), with time to fall asleep and the share of time asleep also improved at every follow-up. The effect shrinks over the year but does not disappear, which a medication taken and then stopped does not match.

In the same 20-trial pool, total sleep time rose by 7.6 minutes, with a confidence interval running from minus 0.5 to 15.7, so the estimate does not exclude zero. The measured change is in how consolidated the night is, not in how long it is. Measured in: 1,162 adults with chronic insomnia across 20 randomized trials, 64% women, mean age 56. This is a null measured at the end of treatment in trials whose active phase deliberately shortens time in bed, so it is partly an artifact of when the outcome was taken. Total sleep time generally recovers as the window is expanded, and this pooled figure does not follow that expansion out. Same source as our sleep-efficiency row, not independent corroboration.

Sleep-hygiene education contributed nothing measurable to remission inside a CBT-I package (incremental OR 1.01, 95% CI 0.77 to 1.32), and the authors classed it as not essential. A separate meta-analysis of hygiene delivered on its own found only small-to-medium improvement from a person's own baseline and significantly worse results than full CBT-I. No measurable contribution inside a package is not the same as no effect at all, and it is the component most readers have already tried.

In adults with obstructive sleep apnea and comorbid insomnia, CBT-I given before CPAP raised initial acceptance of the machine (99% versus 89%) and added 61 minutes to average nightly use compared with usual care, with greater improvement in insomnia severity and in unhelpful beliefs about sleep at six months.

Across 11 randomized trials, internet-delivered CBT-I improved insomnia severity, the share of time in bed spent asleep, subjective sleep quality and time awake in the night, with effect sizes from Hedges g 0.21 to 1.09 held through follow-ups of 4 to 48 weeks. Longer treatment and greater clinical support produced larger effects; higher dropout produced smaller ones.

In 1,711 adults the program beat sleep-hygiene education at week 8 on sleep-related quality of life by a large margin (adjusted difference −17.60) and on functional health (1.76) and psychological wellbeing (2.68) by small ones, with reduced insomnia mediating 45.5% to 84.0% of the change. The trial was run by the program's own team, several of whom disclose salary, shareholding or consultancy ties to the maker, and the comparator was information, not an active treatment, so it establishes that the program beats a leaflet, not that it matches a therapist.

At the one-year follow-up, 56.6% of the automated-CBT-I group (69 of 122) were in remission on the Insomnia Severity Index and 69.7% were treatment responders. The group-by-time comparison against an online education control was significant for insomnia severity, time to fall asleep and time awake in the night, all favoring the program. The large within-group change scores (Cohen d up to 2.32 at one year) also fold in regression to the mean and expectancy, which a control arm exists to subtract, so the remission rate is the figure to read here, not the within-group d. Measured in: 303 adults with chronic insomnia, 218 women (71.9%), mean age 43.3, half with at least one medical or psychiatric comorbidity. A within-group pre-post effect size includes regression to the mean, natural remission and expectancy, all of which a control arm exists to subtract, so the between-group remission rate is the safer figure to read than the within-group effect size.

Guided online CBT-I beat a wait-list on insomnia severity (Cohen d 1.2) and individual face-to-face CBT-I beat it by more (d 2.3). Head to head, face-to-face was better by d 0.9 at every timepoint, and the advantage held on all sleep-diary measures at 3 and 6 months, along with depression and anxiety scores.

Over six weeks of acute treatment, CBT alone and CBT plus zolpidem produced similar response rates (60% versus 61%) and remission rates (39% versus 44%). Over the following six months the best sequence was starting with both and then continuing CBT alone: 68% remission at follow-up, against 42% for those who stayed on medication. It is one trial at one center, and the sequencing comparison is between subgroups, not a fresh randomization, so read it as suggestive; it points the same way as everything else here.

Inside a CBT-I package, relaxation was the one component whose point estimate ran against benefit (incremental OR 0.81, 95% CI 0.64 to 1.02), which the authors called potentially counterproductive. The interval crosses 1, so this is a direction, not a demonstrated harm, and the 2021 AASM guideline separately gives relaxation a conditional recommendation as a stand-alone treatment. It may compete for the limited effort a person can spend on the protocol, or simply be added more often in harder cases, which the model cannot separate.

Against a control that regularized time in bed without shortening it, four weeks of sleep restriction raised evening sleepiness (d 1.17 early, 0.92 late), raised daytime sleepiness scores in weeks 1 and 2, and increased deep-sleep intensity, while pre-sleep arousal fell. Sleep pressure moved only in the arm whose window was cut. It is 56 people at one site, and the mechanism is read from parallel movement in sleepiness and brain-wave measures, not a formal mediation model, but it does establish that the shortening, not the regularity, raises sleep pressure.

A bipolar-specific CBT-I that never restricted time in bed below 6.5 hours produced fewer days in a bipolar episode over six months than psychoeducation (3.3 versus 25.5) and a lower hypomania or mania relapse rate (4.6% versus 31.6%). The authors set the floor explicitly, writing that acute sleep deprivation can bring on next-day hypomanic or manic symptoms. It is a single trial, so read the size of the effect as preliminary; the safety point behind the floor is not.