Thirty-seven RCTs (n=3,957) were included in the review. Fifteen RCTs (40.5%) scored 6 or more on quality. Twenty-eight RCTs (75.7%) adequately described the method of randomisation. Criteria on the baseline characteristics, timing of outcome measures and description of dropouts were met by 50% or more of the RCTs.
Exercise therapy compared to usual care (six RCTs): Two RCTs found a statistically significant reduction in pain post-treatment (WMD -9.23, 95% CI -16.02 to -2.43) and three RCTs found a statistically significant reduction in disability (WMD -12.35, 95% CI -23.00 to -1.69) at short-term follow-up. There were some inconsistencies between the results reported in the text and those reported in the table, but at intermediate follow-up two RCTs found that exercise therapy resulted in statistically significant reductions in disability (WMD -5.43, 95% CI -9.54 to -1.32) and one RCT found a statistically significant reduction in pain relief with exercise therapy (WMD not reported). There was no statistically significant difference in pain at long-term follow-up (two RCTs), but there was a statistically significant decrease in disability in favour of the exercise group (WMD -3.17, 95% CI -5.96 to -0.38).
Exercise therapy showed no statistically significant differences on measures of pain or disability compared to waiting list controls/no treatment (eight RCTs), back school/education (three RCTs), behavioural therapy (three RCTs), spinal manipulation (five RCTs) and psychotherapy (one RCT).
Trials that compared exercise therapy with other forms of exercise therapy could not be pooled, but nine of 11 RCTs reported no statistically significant differences between treatment groups.