Wrist Mobility Biomarkers Predicted Incident COPD in UK Biobank
Key Takeaways
- Among 28,251 older UK Biobank participants, 2.26% developed COPD over a mean 8.5 years of follow-up.
- Lower running duration, slower maximal walking speed, shorter longest walking bout duration, and a lower proportion of walks longer than 8 seconds were independently associated with later COPD.
- All digital gait parameters were significant after age- and sex-adjustment, with running duration showing the strongest association.
- A simplified model showed good discrimination, and in an illustrative 70-year-old man with asthma and one pack-year of smoking, estimated 5-year COPD risk was 7.2% at poor mobility (10th percentile across the four gait biomarkers) versus 1.4% at favorable mobility (90th percentile).
This prospective cohort analysis used UK Biobank accelerometer data linked to National Health Service (NHS) electronic health records, as detailed in the Fernandez et al prospective UK Biobank study of wrist-worn mobility biomarkers and incident COPD. Participants were aged 60–78 years, had no COPD diagnosis at accelerometer collection, contributed at least 5 days of valid data, and wore a dominant-wrist Axivity AX3 device for 7 consecutive days. The investigators extracted 11 digital gait biomarkers with the Watch Walk method and identified incident COPD through International Classification of Diseases, 10th Revision (ICD-10) codes J44, J44.0, J44.1, J44.8, and J44.9, with death and loss to follow-up treated as censoring events and follow-up capped at 9 years. Cox proportional hazards models, backward elimination, and 500 bootstrap resamples were used to estimate associations and assess internal model stability before comparing baseline mobility with later COPD events.
Among 28,251 participants, mean age was 66.7 years and 51.4% were women; 639 developed incident COPD, a 2.26% incidence, over a mean follow-up of 8.5 years, SD 1.3. Those who later developed COPD had poorer baseline performance across all digital gait biomarkers. In minimally adjusted analyses, each 1 standard deviation increment in log-transformed running duration was associated with a 45% lower hazard of COPD, while maximal walking speed was associated with a 37% lower hazard, usual walking speed with a 35% lower hazard, and step-time variability with a 21% higher hazard. The full 15-variable model reached a Harrell concordance index of 0.81 with SE 0.01, while the simplified COPD prediction model using four gait biomarkers and four established variables retained age, sex, smoking pack-years, asthma history, running duration, maximal walking speed, longest walking bout duration, and the proportion of walks longer than 8 seconds, with a Harrell concordance index of 0.80 with SE 0.01 and a mean bootstrap C-index of 0.80. In the illustrative example profile, 5-year COPD risk was 7.2% with poor mobility versus 1.4% with favorable mobility.
These findings were observational associations and should not be framed as causal effects or as evidence of diagnostic accuracy against spirometry-based screening. The authors called for external validation across different age groups, socioeconomic backgrounds, geographic settings, and ethnicities before clinical implementation. They also noted that requiring at least 5 days of wrist-sensor wear could narrow generalizability toward people more willing or able to adhere to wearable use. In the authors’ interpretation, the mobility measures may function as low-burden indicators of elevated COPD risk rather than diagnostic measures.
The authors reported that several wrist-derived, real-world mobility measures were independently associated with later incident COPD in older adults. They also found that a simplified model combining four gait biomarkers with four established variables showed stable internal discrimination. The reported signal linked everyday mobility patterns with future incident COPD diagnosis in this older cohort.
Clinician Questions
Which wrist-derived mobility measures were independently associated with incident COPD in older adults?
Lower running duration, slower maximal walking speed, shorter longest walking bout duration, and a lower proportion of walks longer than 8 seconds were independently associated with later incident COPD in older UK Biobank participants.
How many UK Biobank participants developed COPD after wrist accelerometer monitoring, and over what follow-up period?
Among 28,251 UK Biobank participants aged 60–78 years with no COPD at baseline accelerometer collection, 639 developed incident COPD, corresponding to 2.26%, over a mean follow-up of 8.5 years.
How well did the simplified mobility-based model predict future COPD risk?
The simplified model combined four gait biomarkers with age, sex, smoking pack-years, and asthma history, reached a Harrell concordance index of 0.80 with SE 0.01, and had a mean bootstrap C-index of 0.80, indicating stable internal performance.
What 5-year COPD risk difference was illustrated for poor versus favorable mobility profiles?
In the study’s illustrative scenario, a 70-year-old man with asthma history and one pack-year of smoking exposure had an estimated 5-year COPD risk of 7.2% with poor mobility versus 1.4% with favorable mobility, an approximately fivefold difference.