Sleep Europe 2026: eight findings from the congress

Baker et al. (USA) compared smartwatch recordings with the average of seven nights in postmenopausal women with sleep problems. With a threshold of 60 minutes of wake after sleep onset, a single night misclassified 23.1% (34/147); a window of five nights brought this down to 4.8%.

Rusanen et al. (ESADA network, 24 centres) recorded 702 people with suspected sleep apnea at home with PSG on three nights in a row. For desaturation severity, using the median of three nights instead of the first night moved 11.8% of patients to a different risk group.

Limit: the Baker study had 21 women with seven complete nights, and it was run by a pharmaceutical company.

Source: J Sleep Res 35(S1), Late breaking abstracts, clinical, p. 97; J Sleep Res 35(S1), Oral sessions, p. 47

Bouhamdi et al. (Morocco) calculated hypoxic burden in 984 patients with sleep apnea confirmed by PSG. Of the 880 patients with an AHI of 15 or more per hour, only 199 (20.2%) were in the high hypoxic burden group. Hypoxic burden was moderately correlated with AHI (r=0.652).

The congress also had sessions titled "Beyond AHI" and "Hypoxic load in OSA". Hypoxic burden is the area under the desaturation curve divided by sleep time; its unit is percent minutes per hour.

Limit: a retrospective study from one centre. There is no guideline threshold for hypoxic burden.

Source: J Sleep Res 35(S1), Oral sessions, p. 72

Marchi et al. (France) found an AHI of 15 or more per hour in 37 of 69 patients (53.6%) who had PSG for psychophysiological insomnia. Snoring separated the two groups best (91.7% vs 40.0%).

Wallace et al. followed 851 patients at a veterans' sleep clinic. Comorbid insomnia and sleep apnea (COMISA) was linked to about twice the risk of death compared with insomnia alone (HR 1.90) and sleep apnea alone (HR 2.14).

Limit: both studies are retrospective and each comes from one centre. The Marchi sample is small.

Source: J Sleep Res 35(S1), Poster sessions, p. 509; J Sleep Res 35(S1), Late breaking abstracts, clinical, p. 6

Lu et al. (Australia) followed 168 teenagers for two years. Those who used devices every night before bed fell asleep more than an hour later on average than those who did not; each extra type of device delayed sleep onset by another 14 minutes. Device use was not related to melatonin phase.

Waibl et al. (Switzerland) reported that in 143 teenagers each extra hour of screen time after going to bed shortened sleep by 26.0 minutes on weekdays and 34.2 minutes at weekends.

Limit: both studies are observational; they show that screens and shorter sleep go together, not that screens cause it.

Source: J Sleep Res 35(S1), Oral sessions, p. 31; J Sleep Res 35(S1), Oral sessions, p. 8

Kim et al. (South Korea) fitted 46 people at home with five devices alongside PSG: a phone app, two watches and two rings. All five gave the same sleep stage in only 51.7% of epochs. Most authors work for the company that makes the app being compared.

Concheiro Moscoso et al. (Spain) found a median SpO2 error of 2.83 points for a wrist band in 259 people; the band missed values below 90%. Its heart rate was accurate.

Demey et al. (Belgium, Canada) reported in 15 patients that a ring's oxygen measures agreed with PSG, but it overestimated REM time.

Limit: samples are small and some studies were supported by manufacturers. No abstract tested a phone, a watch and a ring together against PSG on the same night.

Source: J Sleep Res 35(S1), Oral sessions, p. 46; J Sleep Res 35(S1), Poster sessions, p. 222; J Sleep Res 35(S1), Poster sessions, p. 228

Koster et al. (ProPASS consortium) measured sleep duration, regularity and efficiency with a accelerometer worn on the thigh in 14,085 people from six cohorts. Regularity and sleep efficiency were linked to a cardiometabolic risk score independently of sleep duration.

Limit: data from a single point in time; cause and effect cannot be shown.

Source: J Sleep Res 35(S1), Oral sessions, p. 63

Bouchou et al. (France) randomised 79 healthcare workers on 12-hour shifts to a 30-minute nap opportunity in a quiet room between 01:00 and 04:00 or a standard break. After 12 weeks, endothelial function (EndoPAT) was preserved in the nap group and fell in the control group; difference 0.30 (p=0.02). Pulse wave velocity and blood pressure did not differ.

Limit: single centre, small sample and a surrogate outcome.

Source: J Sleep Res 35(S1), Oral sessions, p. 63

Amidi et al. (Denmark) tested the five components of a fully automated smartphone CBT-I app in 478 people with a fractional factorial design. The Insomnia Severity Index fell by 6.6 points on average by the end of treatment. The only component with a significant added effect was sleep restriction (an extra 0.77 points, d=0.29). The trial was preregistered (NCT05561829).

Limit: the app was developed by the authors' team; the effect was slightly smaller at six months.

Source: J Sleep Res 35(S1), Oral sessions, p. 50