CBT for insomnia

CBT-I is cognitive behavioural therapy for insomnia, and it is the recommended first treatment for chronic insomnia in every current major guideline, ahead of medication rather than after it has failed. It is mostly not talking. Its two active components are a sleep schedule calculated from a record of what sleep is actually doing, and a rule about what the bed is allowed to be used for.

The name misleads people twice. It is not relaxation training, and it is not the sleep hygiene list delivered by a professional.

The guideline position is unusually uniform. The American College of Physicians recommends that all adults receive CBT-I as the initial treatment for chronic insomnia, with medication becoming a shared decision only afterwards; the American Academy of Sleep Medicine gives this multicomponent therapy its one strong recommendation; the European guideline says the same in the same order.

The mechanistic reason is more interesting than the endorsements. A hypnotic acts on the state and the therapy acts on what is maintaining it. Chronic insomnia is held in place by a set of responses to bad sleep rather than by whatever started it, which is the argument in the entry on insomnia. A drug suppresses the symptom while it is in the body and leaves the maintaining machinery where it was. CBT-I dismantles the machinery, which is slower, less pleasant for a fortnight, and still working a year later.

The trial that settled the question of combining them shows the difference. Morin and colleagues ran CBT alone against CBT plus a hypnotic; both arms improved during treatment, and then the extended phase separated them, with the CBT-alone group holding its gains best. Adding the drug was not inert. It gave the patient a second explanation for why the night went well.

None of which makes medication a mistake. It is a reasonable short-term decision, sometimes the only thing available, and a taper is the prescriber's business rather than something to improvise alongside a new schedule.

A full course is built from four parts, listed here by what they contribute, which is almost the reverse of their fame.

Sleep restriction repairs a mismatch between opportunity and ability. What gets restricted is time in bed, not sleep, and the name is one of the worst in clinical medicine. Somebody sleeping six hours inside a ten-hour opportunity spends four hours a night awake in a bed. Narrowing the window to something close to the sleep that is actually happening does two things at once: it concentrates sleep pressure, so what remains arrives faster and holds better, and it removes the hours of lying awake that the next component is about. Then the window widens again as the sleep inside it becomes reliable. Spielman and colleagues published the method in the eighties and the direction has not changed.

Stimulus control repairs a learned association. The bed has become a cue for being awake, and the fix is to stop supplying that pairing. It has its own section below, because it is the component people argue with.

The cognitive work repairs the arithmetic and the effort. Not positive thinking about sleep. It targets specific habits that keep arousal up: the calculation at two in the morning of how many hours remain, the certainty that tomorrow is already ruined, the conviction that a particular number of hours is required to function, and the monitoring that Harvey's model puts at the centre, where checking whether sleep is coming is enough to stop it coming. The most useful idea here sounds like a paradox: sleep is the only task that gets further away the harder it is attempted. Effort is arousal, so the clinical move is to reduce trying, which cannot be done by deciding to and can be done by changing what the time awake is spent doing.

Sleep hygiene is in there, and it is the weakest part. It removes obvious obstacles rather than treating anything, guidelines recommend against it as a standalone therapy, and most of the large population who believe they already tried this treatment tried this one piece. The entry on sleep hygiene covers what it is genuinely good for.

Relaxation appears as an optional fifth component in some protocols. It helps some people with the physical end of arousal and it is not what makes the treatment work.

One quantity runs the schedule, and the arithmetic fits on the back of a diary.

Sleep efficiency is total sleep time divided by time in bed, expressed as a percentage. Six hours of sleep inside nine hours in bed is about 67 percent. Good sleepers run in the high eighties or above, and the gap in that example is not neutral dead time; it is three hours a night of rehearsing wakefulness in the place where sleep is supposed to happen.

What makes it the steering quantity is which way it can be moved. Total sleep time is not under anybody's direct control, which is the entire complaint. Time in bed is, completely. So the protocol adjusts the denominator, watches what the numerator does, and widens the window again when the ratio says the bed is reliably producing sleep.

Two counterintuitive things follow. The prescription is built from how much sleep is actually happening rather than how much is wanted, which is why a person asking for eight hours is initially given fewer. And the treatment does not primarily manufacture extra sleep; it removes the waiting, which turns out to be most of what people were suffering from.

The thresholds, the size of each weekly adjustment and the floor below which time in bed is not reduced belong to a supervised protocol rather than to a reader with a calculator. Those numbers, and the order the weeks come in, are in what a course actually involves.

This is the instruction that gets the most resistance, and the resistance is reasonable, so it is worth having the argument in full.

Stimulus control is a conditioning argument and nothing else. A bed is a cue, cues are learned by repetition, and learning of this kind does not consult anybody's intentions. A bed used for sleep comes to produce drowsiness on contact, which is why good sleepers yawn while brushing their teeth. A bed used for lying awake, counting hours and bracing for tomorrow comes to produce alertness, reliably, after enough nights of the pairing. Bootzin's instructions amount to refusing to keep supplying it: lie down when sleepy rather than when it is time, keep the bed for sleep, and if you are clearly awake, get up and go elsewhere until sleepiness returns.

Then the objection. Lying still with the eyes closed does provide some genuine rest, and nobody in this literature disputes that. The problem is that resting quietly while awake in bed is precisely the behaviour that built the association. The rest is real and it is bought with the mechanism of the disorder. Rest can be had in a chair.

Perlis and colleagues explain why it feels so unfair. In their model the conditioned arousal is specific to the sleep context, which is why the same person falls asleep instantly in front of the television and becomes alert on contact with their own pillow. That is not a quirk to apologise for. It is the diagnosis, and it is also the evidence that the association can be re-learned, because what was learned in a particular place can be unlearned there.

Getting up is therefore not willpower and not punishment. The bed can only stop predicting wakefulness if it stops being used for it.

The improvement curve is not a gentle slope from day one, and people who are not told that arrive at the second week feeling ambushed.

The early cost is unavoidable by mechanism. Narrowing time in bed works by raising sleep pressure, and raising sleep pressure means being sleepier. For the first two to three weeks that sleepiness is measurable, it shows up on tests of vigilance and reaction time, and the days feel worse before they feel better. Nothing has gone wrong when that happens. The gains arrive behind the cost rather than alongside it.

Two practical consequences. Naps and early nights borrow from the pressure the protocol is building, so an afternoon nap in week two undoes week two. And the safety side is real: driving is a conversation for the first appointment rather than a discovery in week three.

Which is also why this page describes the mechanism and does not provide a schedule. Sleep restriction improvised alone makes a person sleepier and less safe before it makes them better, and the situations where it should not be attempted are the ones nobody can screen themselves for.

Four situations, and the first one is a line rather than a caution.

Bipolar disorder. Sleep loss is an established trigger for mania, and sleep restriction is partial sleep deprivation administered deliberately for a fortnight. This is not a theoretical worry: when sleep deprivation has been used on purpose as an antidepressant intervention in bipolar depression, a small but real proportion of patients switched into mania or hypomania. The adapted protocol developed by Harvey and colleagues does not merely proceed carefully, it changes the parameters, raising the floor on time in bed and attaching mood monitoring and a written safety plan, and the European guideline lists conditions that sleep loss can worsen as contraindications to the components that involve partial deprivation. A bipolar diagnosis means the schedule is set by somebody who knows about it, and it is the one thing on this page that should never be started on your own initiative.

Untreated sleep apnoea. The two coexist more often than either is diagnosed alongside the other, and the evidence favours treating both rather than queueing them, which requires the breathing to have been identified first. Snoring with witnessed pauses, or unrefreshing sleep at any duration, is a reason to get that looked at before a window is narrowed around it.

Severe depression. Not a contraindication, and CBT-I is used successfully in depressed patients, where it often improves mood as well. What changes is supervision, because a fixed rise time is a task the illness may simply not permit.

Shift work. The fixed wake time is the anchor the whole schedule hangs from, and rotating shifts remove it. Adapted protocols exist; a standard course handed to somebody working nights is the wrong document.

Pregnancy, chronic pain and significant daytime sleepiness from any cause do not rule the treatment out and do change how the window is set. All of which argues for the assessment that comes before the protocol, not against the protocol.

The useful part of this treatment is hard to self-administer and the preparation is not.

• Keep a fortnight of diary. Bedtime, rise time, how long sleep took, time awake in the night, naps, alcohol, a line on the day. It is the instrument every version of this treatment runs on, and it turns a first appointment from a guess into a reading.
• Ask for the treatment by name, since a tablet is the path of least resistance for everybody in the room, and say how wide your current opportunity is. If the time in bed is already tight, restriction is not your problem.
• The test of whether a programme is this treatment is whether it reads a diary, computes a window, and moves that window on what the diary said. Validated digital courses pass and are a legitimate route; breathing exercises and soundscapes do not.

MoodyWonder's Insomnia Check reads onset, maintenance and daytime cost separately, the same three-part shape an assessment starts from. It is a screening-style read rather than a diagnosis.

This is a treatment with a known duration, a known unpleasant stretch in the middle, and effects that outlast it. It is also the one place in the sleep literature where the instinct of the person suffering and the direction of the evidence point opposite ways, which is why it is delivered by somebody else rather than read off a page.