We would like to reply both to Dr. Ernst's letter and to the editorial by Deyo that accompanied our manuscript (2).

We agree, as we pointed out in our paper, that the provision of acupuncture by only 1 practitioner was a weakness of our study and might have implications for generalizability. However, both Dr. Ernst and Dr. Deyo make the same error in interpreting our results and conclusions, that of confusing effectiveness and efficacy. We believe that this is an important distinction.

From our results, it can be clearly seen that acupuncture did indeed have a large effect. It is an effective treatment. Patients, on average, experienced in the region of a 60% reduction in pain from baseline scores in both treatment groups. A clinical improvement of this magnitude is significant and certainly shows a large "effect," which, if this were a drug trial, would probably be a much-sought-after improvement. Such an improvement would also imply that the treatment protocol was not only sound but was delivered competently, credibly, and with conviction in both the acupuncture and control groups. Indeed, that the acupuncturist was able to elicit similar responses to acupuncture from a placebo treatment would tend to suggest that, contrary to Dr. Ernst's comments, he or she was probably quite skilled. The acupuncture treatment protocol used in our study would be very similar to that found in many physical therapy departments throughout the United Kingdom and is therefore generalizable within that context.

In statistical terms, the trial also showed efficacy: The real acupuncture was statistically significantly better than the placebo treatment. However, in terms of our protocol definition of "clinical" improvement over and above placebo, efficacy was not proven. We would stress that many clinicians may feel that our definition of clinical efficacy (not effect) might be rather ambitious (a 30% difference between acupuncture and control treatments) and may feel that the statistically significant improvement we demonstrated is of clinical significance. We would be the first to admit that defining a "clinically significant improvement in pain" is a rather arbitrary decision and may vary from clinician to clinician. However, we are bound to report our protocol definition of effective treatment as our primary analysis.

It would be logical to assume that previous expectation might blunt the apparent benefit of acupuncture, as Dr. Deyo stated (1), and the Editors may have assumed that our patients might have felt that the control treatment would be ineffective. There is no evidence to support this view; rather, our findings show that the reverse was true. The data from the credibility ratings imply that patients probably had a high expectation of a positive result in both treatment groups. This in turn could suggest that such an attitude predisposed patients to achieving a large treatment effect. Our previous research, however, suggests that belief in complementary alternative medicine is not a predictor of outcome (2, 3).

Finally, we agree with Dr. Ernst that it is important to monitor and report adverse effects. Such effects were indeed monitored and are clearly presented in Table 5 of our paper.