ON and OFF Time as a Digital Biomarker

In Parkinson's disease, ON time is the period during which medication is controlling motor symptoms adequately and OFF time is when it is not. As the disease progresses the response to each dose shortens and becomes less predictable, and the clinical goal shifts from reducing symptom severity towards increasing the proportion of the waking day spent ON without troublesome dyskinesia.

That makes ON and OFF time the outcome that medication adjustment is genuinely aiming at. It is conventionally captured with a home diary in which the patient records their state in half hour blocks, a method that has been validated and is used in registered trials, but which is burdensome, depends on the person recognising their own state, and produces recall error over a long recording.

Continuous wrist monitoring offers the same quantity without asking anyone to remember anything.

Wrist worn sensors record movement continuously and algorithms classify each interval as consistent with adequate control, with bradykinesia, or with dyskinesia, producing an estimate of how the waking day was distributed. Dedicated systems developed for this purpose have been evaluated against clinical assessment and against diary data.

The classification is inferential rather than direct. The algorithm does not observe medication state; it observes movement patterns and infers state from them. That inference is trained on populations, so it performs best in the kind of patient it was developed on and less well outside that range. Recordings typically run for several days to capture day to day variability, since a single day is not representative in a fluctuating condition.

The primary use is titration support and trial endpoints in advanced Parkinson's disease, where treatments including device aided therapies are evaluated on their effect on ON time without troublesome dyskinesia. Studies have reported that introducing objective measurement into routine care changes management decisions and improves outcomes, which is a stronger claim than most digital measures can make.

The second use is detecting wearing off, meaning the shortening of benefit before the next dose, which patients often under report because they adapt their routine around it. Measures are reported alongside quality of life instruments such as the Parkinson's Disease Questionnaire, since the point of increasing ON time is what the person can do with it.

Algorithmic state classification is an inference from movement, and movement is affected by much besides medication state, including activity, fatigue and comorbidity. Sitting still is not the same as being OFF, and a device that treats reduced movement as reduced control will misread a quiet afternoon.

The reference standard is itself imperfect: patient diaries are validated but subjective, so agreement between a sensor and a diary is agreement between two imprecise measures rather than validation against truth. Performance also varies by disease stage and by whether dyskinesia is present, and dyskinesia in particular can be misclassified as good control because both involve movement.

• Hauser RA, et al. Parkinson's disease home diary: further validation and implications for clinical trials. Mov Disord. 2004. pubmed.ncbi.nlm.nih.gov
• Griffiths RI, et al. Automated assessment of bradykinesia and dyskinesia in Parkinson's disease. J Parkinsons Dis. 2012. pubmed.ncbi.nlm.nih.gov
• Farzanehfar P, et al. Objective measurement in routine care of people with Parkinson's disease improves outcomes. NPJ Parkinsons Dis. 2018. pubmed.ncbi.nlm.nih.gov
• Ossig C, et al. Wearable sensor-based objective assessment of motor symptoms in Parkinson's disease. J Neural Transm. 2016. pubmed.ncbi.nlm.nih.gov