Electrical Grounding and Sleep Quality in Hospitalised Patients With Cancer

Design. Prospective, randomised, sham‐controlled, participant‐blinded, parallel two‐grouptrial conducted in the oncology ward of New Taipei Municipal TuCheng Hospital.Allocation. Participants were allocated to true grounding or sham grounding by arandomised block design generated with the Research Randomizer web service. The true andsham cords were externally indistinguishable, differing only in whether a conductiveconnection was completed internally; participants were therefore blinded to allocation.Investigators who applied the intervention and measured body voltage were necessarilyaware of allocation.Intervention. Grounding was delivered through a medical‐grade hydrogel patch ofapproximately 4 x 5 cm applied to the forehead and connected by a grounding cord, withthe impedance of the completed circuit limited to less than 100 ohms. In hospital thecord was connected to the earth terminal of the bedside mains outlet; at home it wasconnected to a metal window frame, because domestic outlets in Taiwan do not reliablyprovide a standard earth terminal. Grounding was applied from 21:00 to 06:00 on twoconsecutive nights in the ward, suspended until discharge, and then applied nightly athome for at least six hours until the outpatient follow‐up visit.Body voltage. The alternating potential at the fingertip was measured with a digitalmultimeter while the participant lay in bed, both before and during grounding, on eachin‐hospital study night and once at home.Remote bedroom assessment. Within two days of discharge a researcher conducted a videocall to record the electromagnetic environment of the bedroom using a purpose‐designedquestionnaire, and to teach the participant to measure body voltage and to ground safelyat home.Statistical analysis. The primary between‐group treatment effect was estimated byanalysis of covariance, regressing each follow‐up score on allocation arm with thecorresponding baseline score as covariate. Every model was repeated on rank‐transformeddata as a sensitivity analysis. All tests were two‐sided at the 5% level.