REF-1569 · Ekelund et al., 2019 · BMJ · Paper · PubMed

Does light activity count?

Most vs least active quarter: death HR 0.38 for light, 0.27 for total; 0.52 for moderate-to-vigorous is also adjusted for sitting, so no ranking. Age 40+, 5.8-y median follow-up, measured once: cohort, not proof; illness may drive both.

Main caveat: One measurement, with short follow-up in several cohorts.
moderate evidenceMeasured on non-athletes and recreational people

Authors reprocessed their data to a standard protocol: one vertical axis, 60-second epochs, at least 10 hours on at least 4 days, and intensity cut-offs by counts per minute. Splines showed non-linear associations for every exposure.

The largest step was from the first to the second quarter, a gap of about 5 min/day of moderate-to-vigorous activity or 60 min/day of light activity. The authors' summary is that all intensities count, and that effects are about twice those of self-report studies.

That size could reflect better measurement, as they argue, or stronger reverse causation in short follow-up of older people (our reading).

Protocol

Cohorts: WAT2D, REGARDS, ABC, BRHS, WHS, FHS, NHANES, NNPAS. WHS alone is 16,738 of 36,383, or 46% (our arithmetic), with 4.3 years of follow-up.
Primary model: age, sex, BMI, socioeconomic status, wear time, and mutual adjustment of sedentary time and moderate-to-vigorous activity. Model C added cohort covariates such as smoking and illness.
Quality: Newcastle-Ottawa scores of 8 or 9.
Funding: no specific funding; REGARDS additionally had an unrestricted Coca-Cola grant, with no funder role reported.

Key Results

Exposure, HR vs first quarter (Model B)SecondThirdFourth
Total activity0.48 (0.43 to 0.54)0.34 (0.26 to 0.45)0.27 (0.23 to 0.32)
Light0.60 (0.54 to 0.68)0.44 (0.38 to 0.51)0.38 (0.28 to 0.51)
Moderate to vigorous0.64 (0.55 to 0.74)0.55 (0.40 to 0.74)0.52 (0.43 to 0.61)
Sedentary1.28 (1.09 to 1.51)1.71 (1.36 to 2.15)2.63 (1.94 to 3.56)
Not like for like: only the moderate-to-vigorous and sedentary rows are mutually adjusted. With age, sex and wear time alone (Model A), fourth quarter is light 0.36 (0.28 to 0.46), moderate-to-vigorous 0.35 (0.28 to 0.44). Fuller adjustment (Model C), fourth quarter: total 0.34 (0.29 to 0.41), light 0.44 (0.34 to 0.59), moderate-to-vigorous 0.61 (0.50 to 0.75), sedentary 2.18 (1.61 to 2.95).
Where curves flatten: about 300 cpm (HR 0.34, 0.27 to 0.43), about 375 min/day of light (0.48, 0.38 to 0.63), about 24 min/day of moderate-to-vigorous (0.39, 0.26 to 0.59). CIs are wide at the upper end.
Sedentary: 10 h/day HR 1.48 (1.22 to 1.79); 12 h/day HR 2.92 (2.24 to 3.83), rising more steeply above 9.5 h.
Lag test: excluding deaths in the first two years changed little, with slight attenuation for sedentary time.
Deaths by quarter: 1187 of 9096 in the least active against 214 of 9086 in the most active, 13.0% and 2.4% (our arithmetic). So more than half of all deaths sit in one quarter.
Reporting: Model A prints light third quarter as 0.42 (0.36 to 0.43), and the interval looks misprinted (our reading). A figure legend expands FHS as Women's Health Study, where it is Framingham.

Mechanisms

Measured: none.
Authors' interpretation: the sedentary signal may be the flip side of total activity rather than a separate risk.
Inference: a hazard ratio of 0.64 for about 5 min/day is too large to credit to those minutes alone. The least active quarter probably holds the sickest people, and a two-year lag may not remove that.

Practical Application

Does light activity count? Most vs least active quarter: death HR 0.38 for light, 0.27 for total; 0.52 for moderate-to-vigorous is also adjusted for sitting, so no ranking. Age 40+, 5.8-y median follow-up, measured once: cohort, not proof; illness may drive both.
Good for: a person who is very inactive: moving more at any intensity sits on the steep part of the curve.
Not good for: a claim that 24 min/day is enough for athletes, or a cap for them; few participants reached high intensity minutes.
Not good for: judging a cyclist or swimmer. Hip-worn devices are poor for those (inference), and thresholds came from walking and jogging.

Limitations

One measurement, with short follow-up in several cohorts.
Reverse causation may persist, as the authors state.
Measurement: one axis, absolute thresholds from young adults, two device brands; cycling, swimming and strength work are poorly captured (inference).
Sample: US and western Europe, aged 40 and over, 72.8% women, analysed with sexes combined.
How this entry was checked and changed
Why this level

device-measured exposure is a real improvement, but this is cohort evidence and stays at the observational cap