“Meta-Analysis of Psychotherapy Outcome Studies”

Historical Context

Smith and Glass conducted their research both in response to the lingering criticisms of psychotherapy lodged by Hans Eysenck beginning in 1952 and in an effort to integrate the increasing volume of studies addressing the efficacy of psychological treatment. In a scathing review of psychotherapy, Eysenck had asserted that any benefits derived from treatment could be attributed to the spontaneous remission of psychological symptoms rather than to the therapy applied. His charge prompted numerous studies on the efficacy of treatment, often resulting in variable and conflicting findings.

Prior to the Smith and Glass article, behavioral researchers were forced to rely on a narrative synthesis of results or on an imprecise tallying method to compare outcome studies. Researchers from various theoretical perspectives highlighted studies that supported their work and dismissed or disregarded findings that countered their position. With the addition of meta-analysis to the repertoire of evaluation tools, however, researchers were able to objectively evaluate and refine their understanding of the effects of psychotherapy and other behavioral interventions. Smith and Glass determined that, on average, an individual who had participated in psychotherapy was better off than 75% of those who were not treated. Reanalyses of the Smith and Glass data, as well as more recent meta-analytic studies, have yielded similar results.

Effect Size

Reviewing 375 studies on the efficacy of psychotherapy, Smith and Glass calculated an index of effect size to determine the impact of treatment on patients who received psychotherapy versus those assigned to a control group. The effect size was equal to the difference between the means of the experimental and control groups divided by the standard deviation of the control group. A positive effect size communicated the efficacy of a psychological treatment in standard deviation units. Smith and Glass found an effect size of .68, indicating that after psychological treatment, individuals who had completed therapy were superior to controls by .68 standard deviations, an effect size that is generally classified as moderately large.

Other Findings

While best known for its contribution to research on the general efficacy of psychotherapy, the Smith and Glass study also examined relative efficacy of specific approaches to therapy by classifying studies into 10 theoretical types and calculating an effect size for each. Results indicated that approximately 10% of the variance in the effects of [Page 799]treatment could be attributed to the type of therapy employed, although the results were confounded by differences in the individual studies, including the number of variables, the duration of treatment, the severity of the presenting problem, and the means by which progress was evaluated. The authors attempted to address these problems by collapsing the 10 types of therapies into four classes: ego therapies, dynamic therapies, behavioral therapies, and humanistic therapies, and then further collapsing the types of therapy into two superclasses labeled behavioral and nonbehavioral therapies. They concluded that differences among the various types of therapy were negligible. They also asserted that therapists’ degrees and credentials were unrelated to the efficacy of treatment, as was the length of therapy.

...