We thank Dr. Shekelle for the opportunity to further clarify our methods, data, and conclusions. First, we calculated standardized mean differences using post-treatment mean values and SDs for each group. For the study by Leibing and colleagues (1) only, we used baseline SDs because post-treatment SDs were not reported. We prespecified a comparison of post-treatment scores, which may have been superior to comparisons of between-group changes among trials in which SDs for differences in changes were incompletely reported (2). We recognize that between-group change scores and post-treatment scores may result in different standardized mean differences when there are baseline differences between the groups (as occurred in the Liebing study). To address the possibility that between-group changes and final value analysis methods might result in different effect sizes, we used between-group changes to recalculate our effect sizes for comparisons between sham acupuncture and acupuncture. The trial by Molsberger and colleagues (3) did not report SDs of between-group changes or any statistics that would allow us to calculate these. For the 4-week data from this trial, we assumed a within-subject pretest–post-test correlation of 0.5. For the trial by Mendelson and colleagues (4), the authors appear to have reported 2 different t values in the abstract (t = 0.52) and in a footnote to Table 4 (t = 0.34). Because of our uncertainty over which, if either, of these t values was correct, we have performed 3 different meta-analyses of between-group changes. We used t = 0.34 for the first and t = 0.52 for the second; for the third, we assumed a conservative pretest–post-test correlation of 0.5. Recalculation with our estimated scores of between-group changes resulted in slightly smaller effect sizes (between 0.51 and 0.57) and wider confidence intervals that were still statistically significant.

Second, for the trial by von Mencke and colleagues (5), we used the outcome data from immediately after the end of the first treatment period (that is, a point less than 6 weeks and closest to 3 weeks following the end of treatment). There were no dropouts or crossovers at this measurement point; data shown in Table 12 of the trial report suggested that all 65 patients were available for analysis at this time point. We have used a second independent German translation to confirm the accuracy and appropriateness of the data we used.

Third, we also considered whether trials evaluated credibility of the sham by using a modified Jadad scale, which substituted the credibility testing item for a randomization-stated (yes or no) item. This substitution had the effect of reducing all trial scores by 1 point on the 5-point quality scale. Although we evaluated the credibility testing item and penalized the quality scores of all trials in our sample by doing so, we believe that the validity of this item is unclear and warrants further study.

Last, we believe that our original analyses based on post-treatment mean values and our subsequent analyses based on between-group changes support our conclusion: Current preliminary data suggest that acupuncture may be more effective than inactive controls for providing short-term relief of chronic low back pain.