Dr. Baker stated that insertion of 2 abdominal needles in the sham procedure might have triggered a diffuse noxious inhibitory control mechanism, causing a positive response in the sham group above that for inert placebo. If this did occur, the specific effect size of true acupuncture was underestimated.

We asked a masking effectiveness question after 4 and 26 weeks of treatment. The sham and true acupuncture groups did not differ significantly in the percentage who guessed that they received true acupuncture at 4 weeks but did differ at 26 weeks. If these guesses were related to needling sensation, the groups should have differed significantly by 4 weeks (after 8 treatments). Masking effectiveness is poorly studied. Schulz and Grimes (1) suggested that guesses about treatment received may represent a surrogate for the treatment's effectiveness, an idea consistent with our data. Regardless of assignment to true or sham treatment, mean Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) scores were lower for pain and higher for function in participants who guessed they had received true versus sham acupuncture at 26 weeks (analysis of variance with Tukey post hoc), but not at 4 weeks, when treatment efficacy was not yet at full strength. If this is a generalizable observation, it reveals a conundrum. If guess is driven by outcome, no truly effective treatment can be masked entirely in placebo-controlled trials.

Side effects of acupuncture, mostly from case reports, are associated with negligence on the part of practitioners (2). We reported our serious adverse events, as well as 9 specific symptoms deemed most likely to occur with acupuncture, and provided for open-ended collection of other symptoms. These symptoms did not differ between the 2 acupuncture groups. As with all treatments, properly trained personnel are required. The issue of pneumothoraces is moot because all of the standardized acupuncture points used in our study are located on the extremities.

Drs. Cherkin and Sherman raise the question of high rates of loss to follow-up. While there was a high rate of loss to follow-up in the education group, we targeted our analyses to and based our conclusions on the comparison of the true and sham acupuncture groups, which had identical attrition rates of 25%. This indicates that attrition was noninformative, that is, if it were related to treatment failure, dropout should have had the same effect on efficacy for both groups. Distributions of demographic characteristics for all participants who completed the trial closely tracked the baseline distributions presented in our paper. In addition, imputed data analyses yielded conclusions identical to those of analyses performed using only available data.

Drs. Cherkin and Sherman liken our trial to those where effects are "largely nonspecific." Both for pain and function, the effects of true acupuncture were 33% higher than sham, an estimate of the specific effects. The statistically significant effects we reported are over and above those of the sham group effects. The specific effects of true acupuncture can be classified as "modest," with standardized mean differences of 0.22 in WOMAC pain score and 0.21 in WOMAC function, for difference in improvement from baseline to 26 weeks for true versus sham acupuncture. However, these effects are similar in efficacy to widely used treatments for osteoarthritis. In trials without identified selection bias, nonsteroidal anti-inflammatory drugs have a standardized mean difference of 0.23 in pain score compared with placebo (3) and are modestly better than acetaminophen (4). Hyaluronic acid injections have a standardized mean difference of 0.19 in pain score (79% of the effect of hyaluronic acid injections is potentially accounted for by injecting placebo) (5). Altman writes that polytherapy—concurrently used modestly efficacious treatments—is the superior strategy for treating osteoarthritis of the knee (6). Our study indicates that acupuncture can provide some relief for patients who cannot tolerate conventional treatment and may be appropriate for patients who are benefiting from conventional treatment but want further improvement.