How we label evidence

Sleep advice online tends to sound equally sure of everything. A breathing trick and a decades-tested therapy arrive in the same confident voice. We wanted the opposite: a short, honest signal next to every method, so you can see at a glance how much weight it can bear.

The five labels

How we choose

We ask the same questions in the same order. Is the method recommended in a clinical guideline? How many randomised trials are there, how large, and how similar were the participants to the people who use Baku? Were the results measured objectively or only self-reported? Has anyone repeated them? Some examples from our science page:

• Imagery rehearsal: strong. The American Academy of Sleep Medicine recommends it for nightmare disorder and for nightmares related to PTSD.^[2]
• Insomnia methods: strong. Cognitive behavioural therapy for insomnia, which includes a regular schedule and getting up when you can’t sleep, carries a strong recommendation from the same academy.^[3]
• Slow breathing: good. In a 2023 randomised study, five minutes a day of cyclic sighing lifted mood and slowed breathing more than meditation, but it was a remote, one-month study of 108 mostly healthy adults.^[4] Relaxation is also listed as an option for nightmare disorder.^[2]
• A sound cue during sleep: emerging. In a 2022 study of 36 people with nightmare disorder, those who linked a sound to their rehearsal and heard it again during REM sleep had fewer nightmares after two weeks than those who rehearsed without it. One small study, using a wireless headband, is a beginning, not proof.^[5]
• Lucid dreaming therapy: emerging. It is listed among the options that may be used for nightmare disorder, but the studies behind it are few and small.^[2]
• Dream incubation: exploratory. A 2020 study from MIT showed that a sleep-onset device could bring a chosen theme into dreams in the lab. Fascinating, and offered in Baku as play.^[6]
• NightWare: medical device territory. This smartwatch system, which vibrates to interrupt nightmares, was authorised by the US Food and Drug Administration in 2020 as a medical device. It belongs to a different category, and is not part of Baku.^[7]

How this relates to GRADE

Clinicians use formal systems for this. The best known, GRADE, asks two separate questions. How confident are we in the evidence: high, moderate, low or very low? And how strong is the recommendation: strong, or weak (sometimes called conditional)? Randomised trials start high and are marked down for problems such as study limitations, inconsistent results, indirectness or imprecision; observational studies start low.^[1] The insomnia guideline we rely on used GRADE.^[3]

Our labels are not a GRADE rating. They are a plain-language summary that sits on top of guidelines built with systems like it. What we borrow is its main lesson: how sure we are, and what we recommend, are different questions. That is why something can be well studied and still only “good”, or promising and still only “emerging”.

Checked, and open to change

A clinical advisor, a sleep psychologist experienced in imagery rehearsal, reviews every label before Baku opens to the public. Labels can move in either direction as new studies arrive. A label is never a promise that something will work for you; it is a fair description of what research shows so far. You can see every method and its label on our science page.