Daniel Salzberg

What Makes a Movement Measure a Real Biomarker

By Daniel Salzberg · November 15, 2026
Part of the Healthcare, Measured series · Movement as a Vital Sign, Part 3 of 6

Part 2 ended on a claim I'd rather test than repeat: getting data at home is the easy part, and knowing the number means something is harder. A time on a stopwatch is a number. What anyone can do with it depends on what's been checked.

The FDA describes a biomarker as a defined characteristic that's measured as an indicator of a biological process or of a response to an intervention, and its examples include blood pressure and blood glucose (FDA). Gait speed isn't a lab value, but that wording covers it. What separates a biomarker from a measurement is the evidence behind it, and I find it easiest to ask five questions.

Does it come out the same twice?

That's reliability. A 2024 review in Age and Ageing went through 95 studies of usual and fast gait speed in community-dwelling adults 65 and older. It rated reliability as sufficient for both, with most ICCs between 0.72 and 0.98, but the authors graded the quality of the evidence as low (review). The 30-second chair stand looks similar. The original work by Jones and colleagues found a test-retest correlation of 0.89 in community-dwelling older adults (summary of the evidence). Most of that was a trained person in a clinic or lab. Part 2 covered the small early studies of doing it alone at home.

Does it track something real?

That's validity, and the answer depends on what you say the number stands for. The same review found gait speed predicts later outcomes, though the evidence was low quality. It also found that gait speed in the clinic correlated only weakly with gait speed in daily life, and none of the correlations reached the level the reviewers had set in advance. I think that finding matters a lot for anyone planning home monitoring. A walk down a hallway and a week of walking around the house are different measurements.

The chair stand shows the same thing on a smaller scale. It correlated at 0.77 with weight-adjusted leg press in the original study, but at 0.29 with isometric knee extension in a Hong Kong sample. Whether it's a strength test depends on which strength test you ask.

Can it see change?

Responsiveness is the weak spot. The gait speed review found only three articles on it, and rated usual speed sufficient and fast speed insufficient, both on low-quality evidence. The minimal important change it reports for usual speed runs from 0.03 to 0.12 meters per second. That's a fourfold spread. If a patient's speed moves by 0.05, they may have improved or they may have had a good morning, depending on which study you trust. My read is that many readings over time could narrow that gap, but that's a question for research and not something anyone has shown.

Compared with what?

A chair-stand count of 12 means little without a reference group. Rikli and Jones gathered 30-second chair-stand data on more than 7,000 moderately active adults aged 60 to 94. For women 70 to 74 the middle half scored 10 to 15 stands, and for men the same age 12 to 17 (tables). Those are moderately active people tested in a standard way. As I noted in Part 2, home sit-to-stand times ran well above clinic reference values, so a program that tests at home needs ranges built from home testing.

Does anyone do anything differently?

This one is easy to skip. The FDA says the context of use, meaning how the biomarker will actually be used, determines how much evidence is needed. A number with no threshold and no next step is a chart nobody reads. Some measures have gotten this far. Rikli and Jones also published chair-stand cut-offs tied to keeping physical independence, and a study in people with hip osteoarthritis put the smallest meaningful change at 2 to 2.6 stands. Those give a clinician something to compare against and something to do.

Next

Part 4 is about AI. Models can only work with what they're given, and a note that says "ambulates with assistance" gives them very little. I'll write about what better inputs look like.

Disclosure: I'm VP of Operations at BioMech Health, which works in objective movement measurement. This series reflects my own views and isn't about any specific product.