Psychologic intervention and survival: wishing does not make it so--letter.

D Andersen et al. (1) claimed that breast cancer patients periencing a recurrence survived longer when they had en exposed to a brief psychologic intervention an avere of 11 years earlier. We are concerned that breast cancer tients, clinicians, and researchers might accept these aims at face value, leading patients to mistakenly seek d professionals to mistakenly recommend psychologic tervention to improve the quantity, rather than the qualof patient lives. Regardless of claims made by researchs invested in demonstrating the contrary, no study of ychologic intervention in which survival was an a priori d point and intervention was not confounded with imoved medical surveillance and attention has demonated a survival benefit among cancer patients (2). The rrent study does not alter this conclusion. As stated preously (3) regarding the Andersen et al. parent study (4), aims about survival benefits are medical claims and ould be evaluated using the same standards of evidence hether they arise from psychologic, pharmacologic, or edical interventions. Andersen et al. (1, 4) fail to meet ese standards. Results are based on an extremely small sample and an en smaller number of events. This report isolated 29 tients from the intervention group who subsequently curred, 10 (34%) of whom survived, and 33 from the treatment group, 8 of whom survived (25%). Although e authors present these data as representing a “59% duction in the risks of dying from breast cancer” (1) e numbers are too small to have any real clinical signifnce when considered in absolute terms. Moreover, these mbers are too small to achieve statistical significance in ple analyses or be subjected to multivariate analyses apopriately. The authors do not provide unadjusted results, t report a significant multivariate analysis of study arm, justing for three of nine covariates examined using stepise regression, a procedure that is well-known to overfit ta and generate study-specific results that do not typicalgeneralize to other samples (5). The strategy for selecting variates was not explicit, and these included some variles which the authors claim the intervention affected .g., psychologic distress) but not others (e.g., health