Outcome comparison: what a sleep-apnoea treatment claim has to move

A reader's own report of better sleep and a sleep study's event counts are two different kinds of information. This page lays them side by side, then checks whether a proposed treatment claim has the measurements it would need.

Part 1: subjective report against measured outcome

Every row has the same structure: what a person notices, what a sleep study would record instead, and what each of them can support when a treatment claim is being assessed.
Domain	What the person reports	What a recording measures	What the report alone can support
Night-time breathing	"I snore less." A partner says the snoring stopped.	Snore channel or microphone signal, airflow, and respiratory effort belts recorded through the night.	Snoring is soft-tissue vibration, not an event count. It changes with sleep stage, body position and alcohol, so a quiet night does not establish that the airway stopped collapsing.
Obstructed events	"I wake up gasping less often."	Apnoeas and hypopnoeas per hour of sleep, scored against stated rules, plus the arousal index from the EEG.	A remembered gasp is one event out of a night of them, and waking recall is poor. A reduction in remembered gasping is compatible with unchanged event counts.
Oxygen	"My oxygen number looks better on my ring or finger monitor."	Desaturation index, lowest saturation, and time spent below a stated threshold, from a validated oximeter on a scored study.	A consumer device is not a scored study, and dye-based compounds can shift what an oximeter reports. A changed reading is not automatically a changed saturation.
Sleep quality	"I slept better." A questionnaire score improves.	Total sleep time, sleep efficiency, wake after sleep onset, sleep-stage distribution, and the arousal index.	Global ratings move for many reasons, including expectation. A better rating is a real experience and still not evidence about airway obstruction.
Daytime function	"I have more energy." A sleepiness score falls.	Vigilance testing such as the maintenance of wakefulness test or the multiple sleep latency test, where a protocol is used.	Reduced daytime sleepiness can occur with no change in event counts, and event counts can improve with no change in sleepiness. Neither one substitutes for the other.
The treatment itself	"I feel different when I take it."	Change from a baseline study to an on-treatment study, with a contemporaneous control group and enough nights to absorb night-to-night variation.	Without a baseline measurement, no before-and-after claim is testable. Without a control, the size of the change cannot be attributed to the treatment.

Part 2: does the claim have the measurements it needs?

Diagnosis confirmed by a scored sleep study before treatment started Without a baseline, there is nothing to compare against, and severity is unknown.

A control group run at the same time Symptoms and event counts drift on their own. An uncontrolled before-and-after series cannot separate the treatment from the drift.

Objective primary endpoint, such as events per hour or desaturation index A subjective questionnaire can be a secondary endpoint. It cannot carry a claim about an obstructed airway.

More than one recorded night, or a design sized for night-to-night variation Event counts differ between nights in the same person. A single night either side of a treatment can produce a difference with no treatment effect.

Blinding, or a stated plan for the fact that the compound is visibly coloured A dye that colours urine and sweat tells the participant which period they are in. Trials need to say how they handled that.

Adherence measured rather than assumed A trial that cannot show the treatment was taken cannot interpret a negative result.

Safety reporting that covers the population at risk Sleep-apnoea patients commonly take other medication, and some are at risk from the compound for reasons unrelated to sleep.

Outcome and analysis plan fixed before the data were collected A result chosen after the fact cannot be distinguished from the most favourable number available.

Verdict: nothing selected yet. Select the items that the claim actually has behind it.