Thirty-four RCTs (providing data on 4,108 participants) were included in the meta-analysis. Fifteen trials were judged to contain bias that was likely to impact conclusions. For all other trials any bias identified was judged either unlikely or very unlikely to impact conclusions. No information was reported on the duration of the follow-up period.
Physiological markers
There was a statistically significant reduction in insulin-like growth factor-I (ng/ml) for patients with breast cancer in the physical activity group (MD = -11.9 ng/ml; 95% CI -23.3 to -0.5; four data points from four studies; Ι²=0%) compared with the control group. No other statistically significant effects were found on other physiological markers.
Body composition: There was a small reduction in BMI (MD = -0.4; 95% CI -0.6 to -0.2; 16 data points from 15 studies; Ι²=0%) and body weight (MD = -1.1Kg; 95% CI -1.6 to -0.6; 14 data points from 14 studies) for the physical activity group compared with the control group. But no difference was found for waist-to-hip ratio (MD -0.01; 95% CI -0.04 to 0.02; three data points from three studies; Ι²=0.5%).
Physical function: Physical activity was also associated with benefit for a number of physical function outcomes such as distance walked in six minutes (MD= 29mins; 95% CI 4 to 55; 5 data points from 4 studies; Ι²= 20%) and increased peak oxygen consumption (MD = 2.2ml/Kg/min; 95% CI 1.0 to 3.4; seven data points from seven studies; Ι²=18%).
Depression: There was a large reduction in depression (Beck Depression Inventory) for the physical activity group but with relatively high heterogeneity (MD-4.1, 95% CI -6.5 to -1.8; four data points from four studies; Ι² =47%).
Fatigue: There was a statistically significant reduction in fatigue (Piper fatigue scale) for the physical activity group (MD -1.0, 95% CI -1.8 to 0.01; three data points from two studies; Ι²= 0%).
Quality of life: The physical activity intervention was associated with benefit on mental health (MD 2.4; 95% CI 0.7 to 0.41; two data points from two studies; Ι²=0%), physical function (MD 3.0, 95% CI 0.7 to 5.3; two data points from two studies; Ι²=0%) and social function (MD 3.4, 95% CI 0.4 to 6.4; two data points from two studies; Ι²=0%) on the short form 36 quality of life measure.
Investigating heterogeneity and publication bias
Substantial heterogeneity was identified on physical function for the following outcomes: leg press, left hand grip and right hand grip. Meta-regression analyses suggested high quality studies reported smaller effects on leg press. For left hand and right hand grip, age of patients and year of publication impacted effect estimates. For most other outcomes there was low heterogeneity.
Small study effects/publication bias was identified for the waist-to-hip ratio outcome. For all other outcomes no such bias was found.