A Meta-Analysis of Psychosocial Interventions for Cancer Patients Gone Awry

Abstract

These claims could have important policy implications and be used to overrule patient preferences or restrict services made available to cancer patients. Our review of the meta-analysis, however, found it bewildering in terms of search strategy, basic definitions, coding and computational errors, and summary statistics and conclusions. Overall, it was simply not credible. Reports of meta-analyses should be sufficiently transparent to allow independent evaluation of how studies were selected, coded, combined, and analyzed. We were struck by the sparse details provided in the paper and some obvious omissions from the studies included. For example, a number of missing studies from nursing journals [e.g., 2, 3] would have raised the number of CBT interventions delivered by nurses and social workers, which could have influenced answers to the declared research question of whether efficacy varied with who delivered the intervention. “Psychosocial intervention” was never defined. Whatever definition was employed resulted in inclusion of a melange of studies (i.e., studies that evaluated the effects of providing a tour of the cancer center to a mixed sample of patients [4], and a cosmetic class with free cosmetics to BC patients [5]). Tours of a cancer center and provision of an audiotape to patients of a consultation with their oncologist were further coded “psychoeducational.” Zimmerman et al.’s use of this term is overinclusive. Psychoeducational is more commonly reserved for interventions providing information about psychosocial issues, such as stress management, coping strategies, communication, or impact on relationships [e.g., 6]. Outcomes relevant to the research were not defined. A diversity of outcomes, ranging from satisfaction with getting a tour to psychological distress to returning to work, was considered comparable and collapsed within and across studies. Summary effect sizes that ignore such diversity in outcome variables and discrepant findings among them cannot be clinically meaningful. In addition, the definition of BC versus heterogeneous cancer samples was flawed. Based on the definition used by Zimmermann et al., a sample with even one non-BC patient was considered a heterogeneous group and would have been coded equivalently as a sample of almost all BC with 1–2 other cancer patients. Because studies almost never distinguish specific effects on BC patients from effects on the rest of the heterogeneous samples, there was no way of determining, as Zimmermann et al. intended, whether BC patients had better outcomes when participating in interann. behav. med. (2009) 37:94–96 DOI 10.1007/s12160-008-9075-2