What is the best way to measure gait in a clinical trial?

Choose the method from the endpoint rather than from what the technology can do. Define the construct, then select an instrument that is fit for purpose in the trial population, applies one identical protocol at every site, and produces auditable source data. FDA guidance asks sponsors to verify that a digital technology measures the parameter accurately and to validate that it assesses the intended clinical characteristic in that population.

Last reviewed .

Start from the construct, not the instrument

The first decision is what the endpoint is meant to represent. Walking capacity, dynamic balance, endurance, and lower-limb strength are different constructs, and gait measurement can serve any of them badly if the mapping is left implicit.

FDA guidance on incorporating clinical outcome assessments into endpoints asks sponsors to describe the type of assessment, the concept of interest, and the specific measure used, and states that endpoints should reflect an aspect of health that is meaningful to patients and be capable of supporting an inference of treatment effect. It classifies performance outcome measures separately from patient-reported, observer-reported, and clinician-reported measures.

Verification and validation are distinct

FDA guidance on digital health technologies for remote data acquisition draws a distinction that is frequently collapsed in vendor discussions.

  • Verification — objective evidence that the parameter the technology measures is measured accurately and precisely.
  • Validation — objective evidence that the selected technology appropriately assesses the clinical event or characteristic in the proposed participant population.
  • Usability evaluation — identifying and addressing use errors or difficulties that participants and site staff may encounter.

A system can be well verified and still not validated for the population in a given trial. The guidance frames the whole question as whether the technology is fit for purpose for that specific investigation, and recommends discussing endpoint plans with the relevant review division.

It also distinguishes measurements that replicate an existing clinical measurement, where a new endpoint justification may not be needed, from novel endpoints, which require justification of their own. Verification and validation are expected in both cases.

Standardising across sites

In a multi-site trial the dominant threat to a functional endpoint is usually not instrument precision but protocol drift between sites. Anything left to local discretion becomes a source of between-site variance.

  1. Fix the walk distance and whether acceleration and deceleration are included in the measured segment.
  2. Fix the pace instruction and the exact wording used to deliver it, since usual pace and maximum pace are different measures.
  3. Fix the number of trials and the rule for combining them.
  4. Fix footwear, surface, and the handling of assistive devices, and record any deviation.
  5. Remove manual timing and local scoring where possible, so the measurement does not depend on which coordinator ran the visit.
  6. Process all sites through one identical pipeline rather than aggregating locally derived numbers.

Retaining the raw capture rather than only the derived value is what makes a functional endpoint auditable. A number produced by a stopwatch cannot be re-derived; a recording can be reprocessed.

Choosing among the practical options

Timed walk tests are inexpensive, universally available, and carry the largest body of interpretive data, but depend on manual timing and yield a single parameter. Instrumented walkways give richer output at a fixed installation per site. Wearables capture gait outside the clinic but introduce adherence and wear-position variability. Camera-based markerless analysis needs no installation and retains the raw capture, with published agreement against reference systems strongest for spatiotemporal parameters and weakest for out-of-plane kinematics.

No option is best in the abstract. The selection is a fit-for-purpose argument about a particular endpoint in a particular population.

Where CurveAssure fits

CurveResearch is built for the multi-site standardization problem: the same video-based assessments recorded at every site, in clinic or at home where the protocol allows, processed through the same pipeline rather than scored locally. Digital-only access is available for protocols that need the sit-to-stand, upper limb motor, patient-reported outcome, and case report form components without shipping hardware.

CurveResearch is for research use only. It is not for use in clinical diagnostic procedures or as a basis for clinical decision-making. Device descriptions covering the computer vision motion analysis system, the calibration floor marker, the grip strength system, and data handling are available to authorized research partners.

References