We thank Drs. Holman and Lorig for their interest in our paper. Regarding their concern about the effect size in 1 study, our calculated effect size is correct according to our methods because it is based on the follow-up means for pain and function in the 2 groups. The effect size listed by Drs. Holman and Lorig in their letter is a "difference of difference" calculation that adjusts for baseline differences in pain and function between the 2 randomized groups in their study. Although this is 1 valid method to analyze the results of an individual article, we prefer not to combine "difference of difference" estimates with follow-up mean estimates in our meta-analyses because of reports that such an approach can increase susceptibility to bias (1). On a practical level, combining just those studies that contained "difference of difference" estimates would decrease the number of studies that can be included in a pooled analysis. In most circumstances, effect sizes that are calculated by using the "difference of difference" method or by using the follow-up means do not substantially differ. In this 1 particular study, they do. However, our pooled result is not very sensitive to which effect size is used from this study or similar studies; use of the "difference of difference" effect size preferred by Drs. Holman and Lorig only changes the pooled effect size for pain by 0.01. If we substitute a "difference of difference" effect size for all 6 studies where it is possible to do so in our pooled analysis of 14 studies, the effect for a decrease in pain diminishes to statistical insignificance (effect size, –0.05 [95% CI, –0.12 to 0.03]).

As for the studies that Drs. Holman and Lorig say we omitted, we must emphasize that our study question of interest was the effect of chronic disease self-management programs on patients with osteoarthritis. We identified all of the studies listed in their references and rejected only those that enrolled patients with mixed musculoskeletal conditions if the proportion of patients with osteoarthritis was unacceptably low (in the 2 excluded studies they mention in their letter [2, 3]), approximately half of the patients had osteoarthritis compared with 77% of those in the study that we did include in our analysis [4]). Readers interested in a synthesis of evidence regarding self-management programs for patients with mixed diagnoses of arthritis are referred to the meta-analysis by Warsi and colleagues (5), which reported a pooled effect size that was very similar to our result (effect size for pain, 0.12; effect size for function, 0.07).

Drs. Holman and Lorig question the validity of assessing a diverse set of studies about educating patients to help manage their osteoarthritis. This misses 1 of our study's key points: There is no agreement on what constitutes a self-management program. Therefore, we included a broad array of studies and used meta-regression to attempt to identify components of particular significance (a method we used to identify particularly effective components of interventions to improve screening and prevention [6]). Unfortunately, our attempts were unrevealing. Our results were also echoed by Warsi and colleagues (7) in another analysis in which meta-regression could not discern meaningful differences in the effectiveness of self-management programs as a function of a large number of different program components. There is simply insufficient evidence to conclude which components of a self-management program are most important in terms of effectiveness. We disagree with Drs. Holman and Lorig that longitudinal studies are "valid methods" to assess the effectiveness of interventions for a chronic disease with a variable clinical course; however, we do agree that there are important outcomes other than pain and function, which is why we stated this as a limitation of our analysis. Unfortunately, we could not include other outcomes because they were reported too infrequently and too variably to justify inclusion. Furthermore, we did not conclude that self-management programs have "no clinically beneficial effect" for older adults with osteoarthritis, only that there are no data to suggest those benefits include clinically important improvements in pain and function.

Future research is needed to determine which clinically and financially important outcomes are reproducibly improved by chronic disease self management programs for older adults with osteoarthritis, and which components are necessary in order to achieve these improvements.