Stigma in mental illness

medicine. However, there is a barrier in the path of meeting this challenge, apart from the fact that this field of evaluation is both under-resourced and difficult. That barrier is reviewer bias. Systematic review is a powerful tool but it is not the end of the story. How the evidence is graded for methodological quality, synthesized into encapsulating statements and used to draw conclusions and make recommendations, demands a high level of rigour and an irreproachable freedom from bias. Ernst points to one back pain triall (an undoubtedly 'random' example) in which one treatment arm consisted of a form of manipulation attributed to chiropractic. This trial showed little difference between any of three interventions and Ernst suggests that it 'has contributed to increasing doubts about whether chiropractic is helpful for acute uncomplicated low back pain in a clinically relevant way'. Readers, however, may not be aware that, at the time of the Royal College of General Practitioners' Clinical Guidelines and Evidence Review of Acute Back Pain2, there were 36 such trials of manipulation (many involving chiropractic) and 53 reviews of this evidence. Taking into consideration relevance, methodological quality and outcomes, these national clinical practice guidelines reflect little doubt, let alone increasing doubt, about the effectiveness of manipulation, chiropractic or otherwise. How are clinicians and patients to deal with such inconsistencies? I suggest that, from now on, they ignore selective reference to individual trials. In terms of evidencebased practice, reference to one trial, unless it is of very high methodological quality and shows a clear superiority of one treatment over another, will reflect only the professional bias of the writer. It does not cast doubt on, confirm, or indicate any significant change in a large body of evidence.