Wallmann-Sperlich asked 614 German desk workers two questions: how they spend their working hours, and how they would like to. They reported sitting for 73.0 per cent of the time, standing for 10.2, walking for 12.9 and doing something physically demanding for 3.9. Asked what they wanted, the same people said 53.8 per cent sitting, 15.8 standing, 22.8 walking and 7.7 physical. Every gap was significant at p below 0.001. The two widest are sitting, at 19.2 percentage points, and walking, at 9.9.
Leppe-Zamora pooled 18 trials with 1,164 office workers on computer prompt software alone, the cheapest intervention in this literature. Measured by accelerometer, it produced 1,029.99 more steps a workday, at a confidence interval of 815.97 to 1,244. That is the largest pooled step figure here and it comes from software that does nothing but interrupt.
André Mamede tested a richer version on 298 employees of the Rotterdam municipality, randomised by office location against an active control rather than against nothing. During the five-week gamified phase, with a digital app and social features, participants took 634 more steps a day than the control group, at p = 0.01. Then the phase ended. A second five weeks of physical nudges in the building produced nothing (p = 0.76), and neither did the follow-up (p = 0.88). Missing accelerometer data also rose to 39.6 and 59.6 per cent by follow-up, which is its own warning.
Daniel Brown randomised 94 office workers to a control group, a lunchtime walk through trees and maintained grass, or a lunchtime walk through housing estates and industrial areas. At peak, the nature group gained 745 steps a day, the built-route group 374 and the control 217. The nature group was the only one to reach the increase the study had aimed for. Adherence across the whole trial was 42 and 43 per cent, which Brown reports as the reason the intervention did not do more.
Phaswana and Gradidge gave 44 South African office workers a height-adjustable desk and kept 18 as controls. Light physical activity rose by 4.14 minutes a day in the intervention group and fell by 14.71 in the control. The standard deviations are around 51 minutes and the comparison comes out at p = 0.691, so the direction is all there is.
Only Mamede and Brown randomised, and both used accelerometers. Wallmann-Sperlich’s 614 workers reported their own hours. Phaswana’s 62 split 44 to 18, which is an unusual allocation for a trial that size.