The researchers searched the medical literature up to October 2004 to find randomized, controlled trials that compared exercise therapy with placebo, no treatment, conservative management, or another exercise group. They classified the types of exercises that made up the interventions. They also looked at several characteristics of the exercise therapy interventions, including whether they were delivered as group or as individual programs and whether they were supervised or done independently without supervision. They then combined data from the trials to explore which exercises reduced pain and improved physical function.