Zimmermann, T., Heinrichs, N., & Baucom, D.H. (2007). “Does one size fit all?” Moderators in psychosocial interventions for breast cancer patients: A meta-analysis. Annals of Behavioral Medicine, 34, 225–239.

To illuminate the moderators of the effect of psychosocial interventions and better understand the variability of these effects in patients with cancer; to test the hypothesis that cancer type, intervention type, and interventionist can moderate intervention effect

• Databases searched were PsycINFO, PSYNDEX, MEDLINE (1980–October 2004). Search keywords were cancer, neoplasm, carcinoma, oncology, breast cancer, psychological therapy, psychotherapy, intervention, counseling, group support, peer support, relaxation, imagery, coping skills training, cognitive therapy and psychological support. Investigators also did some manual searching, using references from retrieved articles.
• In cases in which the literature reported incomplete outcomes, investigators communicated with the authors of the studies to get the data. Some authors provided the necessary data.
• Literature included in the meta-analysis reported, in English or German, on randomized controlled trials (RCTs) and provided complete documentation of outcome measures.
• Literature excluded from the meta-analysis did not report on an RCT or did not include complete data, including the sample sizes of the treatment and control groups; group means; standard deviations or values of t tests, F tests, and chi-square tests.

Initially, investigators retrieved 127 articles, of which 46 were eliminated because they did not report on RCTs. Four of the 127 articles were unavailable. Of the remaining 77 reports of RCTs, the data in 26 were incomplete. Communication with the authors of studies with incomplete data resulted in obtaining complete data for five studies.

• The final number of studies included in the meta-analysis was 56. Sample range was 19–638.
• Meta-analysis, across studies, included a total of 6,419 patients. Mean patient age was 52.4 years (SD = ±4.7 years), and the age range was 40–65 years. The majority of study participants  (89%) were female.
• Studies in the meta-analysis were conducted in the United States or in multiple European countries. Thirty-four studies involved a homogeneous group of patients with breast cancer, and 22 studies involved mixed diagnoses.

• Across all studies, effect size (ES) d = –0.47–2.66; overall average, ES d = 0.56.
• Types of interventions were as follows. Some studies involved more than one treatment group.
  • Psychoeducational (EDU): 8 studies (ES/d = 0.53).
  • Cognitive behavioral therapy (CBT): 24 studies (ES/d = 0.19).
  • Supportive therapy (SUP): 15 studies (ES/d = 0.13).
  • Relaxation (REL): 12 studies.
• Moderator effects were as follows.
  • The majority of studies were led by psychologists (n = 21) or nurses (n = 11). Psychologist-led interventions had a higher ES (d = 0.30) than did those led by nurses (d = 0.15) (p < 0.001).
  • Nature of the control group in studies was identified as active if active interventions of some type were delivered. The control group was identified as passive if controls received usual care or were wait-listed. Passive control groups had higher ES (d = 0.34) than did those with active controls (d = 0.11) (p < 0.001).
  • Formats of the intervention were individual, group, couple, self-help, and patient with family-member. Comparison between group and individual formats showed higher ES for individual interventions (d = 0.30) than for group interventions (d = 0.19) (p < 0.001).
  • Timing of the intervention was reported in 82% of studies. Timing options were directly after diagnosis or surgery, during or following chemotherapy or radiation therapy, and months or years after initial diagnosis. Interventions conducted directly after diagnosis or surgery (d = 0.33) had higher ES than did those during treatment (d = 0.18) or those conducted months or years after diagnosis (d = 0.16) (p < 0.001).
  • Patients with early-stage (I–III) disease had higher ES (d = 0.32) than those with advanced disease (d = 0.13) (p < 0.01).
  • For EDU, moderate ES (d = 0.73) resulted if the intervention had been conducted by nursing or medical staff rather than by psychologists. Psychologist-delivered ED resulted in small ES (d = 0.27).
  • CBT resulted in moderate ES (d = 0.40) if conducted by psychologists. CBT resulted in no effects if nurses or social workers delivered the intervention.
  • Authors found no difference in the results of interventions delivered by different SUP interventionists.

• Psychosocial interventions of various types can have a positive effect on patients with cancer.
• Psychoeducational interventions appear to have the greatest effect, and this effect is higher when the education is delivered by professionals with medical expertise than when delivered by others.
• Interventions appear to be most effective in cases of early-stage disease and when delivered early in the phase of care.
• Individualized interventions appear to have an effect that is greater than interventions delivered in groups.

• The quality of individual studies included was highly variable.
• Some studies provided actual data on significant findings only, although the studies had measured several outcomes. As a result, the ES reported in this meta-analysis may be somewhat overestimated.
• Long-term effects of intervention types are still unclear because most studies did not include long-term follow-up.

Nurses are particularly suited to providing psychoeducational and educational interventions for patients, and these types of nurse-led interventions can have a positive effect on patient outcomes. Psychologist-delivered CBT interventions seem to be more effective than nurse-delivered CBT interventions. This finding suggests that, if nurses are to provide CBT, the nurses must develop significant expertise.This finding may also suggest the importance of interdisciplinary approaches to providing psychosocial interventions. This meta-analysis demonstrated that moderators have a significant effect on intervention effectiveness; therefore, future studies should provide information about potential moderators.