The night is where Parkinson's disease does some of its quietest damage. A person who moves reasonably well at four in the afternoon can be stranded on their back at four in the morning, unable to turn, waking a partner who has learned to sleep with one ear open. Families know this terrain intimately. So does the wearables industry, which now sells the promise of capturing it: a small accelerometer, a week of nights, a dashboard of numbers. The appeal is obvious. Sleep quality is one of the few things in Parkinson's that a machine might plausibly track without anyone having to remember, estimate, or minimise it in a clinic appointment six months later.
Which makes the findings of a systematic review published in npj Parkinson's Disease on 9 July 2026 worth reading slowly. A team led by Manon Mijnsbergen at Leiden University Medical Center, with colleagues from the sleep-wake centre at SEIN in Heemstede and from Amsterdam UMC, gathered 70 studies that used movement sensors to assess nocturnal activity and sleep in people with Parkinson's. Their summary of the field is not a quibble about statistics. They report that these studies are methodologically highly heterogeneous and lack standardization, and they frame their own review as an attempt to inform standardization protocols for research and for clinical use. In other words: seventy attempts to measure the same night, and no shared way of doing it.
The consequence shows up most plainly in what the authors could not do. They ran meta-analyses where the data allowed, and at group level those analyses indicate that actigraphy can distinguish people with Parkinson's from healthy controls. But when it came to intervention studies, the question every family actually cares about, whether a change in treatment or routine shows up as a better night, no meta-analysis was performed, because heterogeneity between the studies was too high. The authors state that sensitivity to detect change in wearable-derived measures remains unclear. A measurement that can tell two groups apart on average is not the same thing as a measurement that can tell you whether your own last three months got better or worse.
This matters for the dementia question specifically, and it matters because of how easily the leap gets made. Night-time measures are increasingly treated as a window onto how Parkinson's is progressing, including the cognitive side of it, and related work is moving fast: a separate study published in npj Digital Medicine on 13 June 2026 described ActiTect, a machine learning pipeline built on standardized actigraphy to screen for REM sleep behaviour disorder, developed and validated across cohorts contributing more than 1,809 nights of recording on a single device type, the Axivity AX6. That is what standardization looks like when someone insists on it. The Leiden review is a reminder of how much of the existing literature was built without it, and therefore how much weight any single wearable-derived sleep number can honestly bear when it is offered as a proxy for progression or for future cognitive change.
The review does not end in despair, and it should not be read that way. Its authors set out what reliable, standardized assessment would require: at least seven nights of monitoring, appropriate sensor placement, and validated algorithms built for Parkinson's rather than borrowed from general sleep research. They also note that studies comparing wearables against polysomnography, the full overnight laboratory recording, show actigraphy can accurately estimate sleep parameters for individual people with Parkinson's. The tool is not broken. The instruction manual has seventy competing versions.
For a family, the practical translation is narrower than the marketing. A wrist device counting a person's restless hours is producing a real signal about something real. What it is not yet producing, on the evidence this review assembled, is a number that can be compared confidently against last year's number, against the person in the next chair, or against a threshold that says anything about what is coming. Anyone weighing what their own device readings mean is weighing a question for their neurologist, who knows the person as well as the data.