Much Ado? A Reply to Hofmann and Smits

    loading  Checking for direct PDF access through Ovid


I welcome the opportunity to address the important issues regarding the conduct of meta-analysis, and I thank Hofmann and Smits for their comprehensive analysis of the differences between our 2 studies. Having read a preprint of the Hofmann and Smits (in press) paper, I am pleased to see the extent to which their data replicated those reported previously by Norton and Price (2007). This convergence greatly strengthens the confidence we can place in the overall effect of cognitive behavior therapy (CBT) on anxiety.The apparent point of contention, it would seem, prompting their Letter to the Editor (Hofmann and Smits, 2008b), relates to the slight differences in the reported comparable efficacy of CBT across diagnoses. Norton and Price (2007) reported that effect sizes for generalized anxiety disorder (GAD) and post-traumatic stress disorder (PTSD) were significantly larger than those for social phobia, while no other differences among diagnoses were observed. Hofmann and Smits reported that, when analyzing data from continuous measures, the effects of CBT for Acute Stress Disorder were greater than for all other diagnoses save obsessive-compulsive disorder (OCD), that CBT for OCD had a larger effect than did CBT for panic disorder, and that no other diagnostic groups showed any differences from each other. The data from both the Norton and Price (2007) and the Hofmann and Smits (in press) studies show essentially the same result: There are few differencesbetween anxiety diagnoses in their response to CBT. It, consequently, seems nonplussing to consider quibbling over which study's nonsignificantly-different-rank-ordered-list is more valid. However, I would like to address some of the specific issues raised in the Hofmann and Smits Letter to the Editor 2008.Hofmann and Smits indicate that only 13 of the 27 high-quality studies included in their analysis were included in Norton and Price (2007). It is true that in a small number of those cases, an eligible study was not included due to an apparent search oversight on the part of this author. In other instances, however, papers were excluded from the Norton and Price (2007) study as they did not meet the strict inclusion criteria. To wit, one of the studies included by Hofmann and Smits (in press) but omitted from Norton and Price (2007) did not require that all participants meet full diagnostic criteria for Acute Stress Disorder, which was one of the Norton and Price—and presumably Hofmann and Smits—study inclusion criteria. This is not meant as an attack on Hofmann and Smits’ study; rather, it was meant to highlight that it is likely that nearly every meta-analysis has inaccurately included or inaccurately excluded a small number of studies. Other discrepancies between our reference lists were a function of some studies not published in a peer reviewed journal (another Norton and Price inclusion criterion), or were published after the completion of our study. I have not taken inventory to check if Norton and Price (2007) included studies which should have met inclusion in, but were not reviewed by, Hofmann and Smits (in press). That 2 meta-analyses have somewhat different reference list does not indicate that either search was not comprehensive.Second, the letter by Hofmann and Smits states that “Norton and Price limited their search to only peer-reviewed studies written in the English language that appeared in PsycInfo or Medline.” This is simply a false assertion.

    loading  Loading Related Articles