Eight studies were included (n=427). These included five RCTs (n=335) and three non-randomised controlled studies (n=92). Jadad quality scores ranged from 1 to 3 out of 5. One study reported use of an adequate randomisation method. None reported participant blinding or adequate allocation concealment. Only two studies assessed compliance; rates were 77% and 85%.
There was no statistically significant difference between walking interventions and controls in lumbar spine bone mineral density (WMD 0.007 g/cm2, 95% CI -0.001 to 0.016, p=0.09). No significant heterogeneity was found. The increase in femoral neck bone mineral density in walking groups just reached statistical significance (WMD 0.014g/cm2, 95% CI 0.000 to 0.0028, p=0.05). Significant heterogeneity was found (I2=51.4%).
There was no statistically significant difference between walking interventions and controls in lumbar spine or femoral neck bone mineral density after excluding non-randomised studies, studies with high attrition rates and studies that used nutritional supplements, or when studies were grouped by methods used to measure bone mineral density.
The funnel plot for studies of lumbar spine bone mineral density was asymmetrical, which suggested potential for publication bias. The funnel plot of femoral neck bone mineral density was symmetrical.