Quante · Nature and science of sleep 2018 · prospective comparative validation study · n=35

Actigraphy-based sleep estimation in adolescents and adults: a comparison with polysomnography using two scoring algorithms.

Cited 305 times in the scientific literature.

Level 3 - non-randomized controlled study

Prospective method-comparison validation study against a reference standard (polysomnography)

PubMed 29403321 · doi:10.2147/NSS.S151085 · record verified 2026-08-29

What was done

Authors compared sleep-wake estimates from two wrist actigraphs (GT3X+ versus Actiwatch Spectrum [AWS]) to in-home polysomnography (PSG) in 35 healthy volunteers (13 school children, 22 adults; 46% male). GT3X+ data were scored using two algorithms (Sadeh and Cole-Kripke). Twenty-two adults wore both devices simultaneously for at least 5 consecutive days and nights; concurrent actigraphy and PSG were recorded over a single night in 12 adults and 13 children. Analysis used intraclass correlation coefficients (ICCs), epoch-by-epoch comparisons, paired t-tests, and Bland-Altman plots.

What was found

Both actigraphs showed 0.81–0.86 accuracy for sleep-wake estimation compared to PSG. For GT3X+, Cole-Kripke was more sensitive (0.88–0.96) to detect sleep but less specific (0.35–0.64) to detect wake than Sadeh (sensitivity 0.82–0.91, specificity 0.47–0.68). Total sleep time on GT3X+ had good agreement with PSG (ICC = 0.64–0.88), while wake after sleep onset agreement was poor (ICC = 0.00–0.10). In adults, GT3X+ with Cole-Kripke matched AWS closely (mean bias 3.7 ± 19.7 minutes for total sleep time, 8.0 ± 14.2 minutes for wake after sleep onset).

Why it matters

Actigraphy provides dependable estimates of total sleep duration across standard devices, but researchers and clinicians cannot rely on it to detect nocturnal wake episodes accurately, and algorithm choice markedly alters inter-device agreement.

Limits

The sample size was small (n=35 total, only 25 with concurrent PSG) and restricted to healthy volunteers. PSG comparison was conducted over only a single night, and wake-detection specificity was uniformly poor across both algorithms and devices.

Cited by