Thirty-one controlled trials were included; six were quasi-randomised trials (one also collected preference information), seven had a preference design, and 18 collected preference information. Thirty trials were randomised – eight had adequate allocation concealment. Seven trials reported blind outcome assessment.
The questions used to elicit preferences varied across trials and were poorly reported. Seven trials reported statistical tests of differences in patient characteristics between those with and those without preferences. Four of these noted that the group with preferences had different characteristics; less education, worse initial measure score, and more or fewer years with pain.
There was a tendency for higher attrition in groups of patients who had no preference, at six months, but not earlier. About three quarters of patients declined randomisation and chose non-trial treatment; around 20% of patients chose acupuncture when offered several treatments.
Ten trials reported the effects of preferences on outcomes; only one found a statistically (but not clinically) significant difference.