To the Editor: I am concerned as to the validity of the recent review by Juan R. Bustillo, M.D., and colleagues (1) on the psychosocial treatment of schizophrenia. First, this review did not follow the procedures currently considered standard in research synthesis. The authors’ literature search appears reasonably comprehensive but may not have been as exhaustive as might be expected in a high-quality journal. For family therapy, for example, the authors failed to mention the trial by Buchkremer et al. (2), which resulted in negative findings. More important, the authors did not appraise—or did not clarify how they appraised—the methodological quality of the individual studies identified; there are so many different ways to proceed, and the validity of the review naturally hinges on which studies the reviewers chose to include (3).

Furthermore, the review did not attempt a meta-analytic pooling of the results. Quantitative synthesis is not a necessary condition for a scientific overview. However, failing to do this in meta-analyses in which it is possible renders the review vulnerable to a selective reporting of results from trials that are consistent with the reviewers’ preconceived notions. It also deprives the review of chances to test for heterogeneity and publication bias; this is one of the cardinal requirements of research synthesis in medicine (4).

All this may partly explain my last concern with this review, namely, that the conclusions drawn do not seem to follow logically from the findings in the trials reviewed. In the text, the authors noted the remarkable lack of relapse prevention in the most recent trials of family therapy but recommended it for the "majority of persons suffering from schizophrenia" (1, p. 163). The authors attributed the negative findings of the more recent trials to low event rates, but systematic review over time does not warrant such an interpretation (5). Regarding social skills training, the authors wrote, "There is little evidence that this learning [with use of the basic model] translates into improved social competence in the community" (1, p. 167), and noted only modest benefit, such as in two of six or three of 10 areas of social functioning that they examined, with other, more intensive modes of social skills training. If these statements are true, then how could the authors conclude by writing, "Patients with schizophrenia can clearly improve their social competence with social skills training" (p. 163)?