How big is big enough? Thinking about contralateral prophylactic mastectomy.

When evaluating the benefit of a therapy, we must consider the effects of uncertainty and the magnitude of the gain or loss, having decided on a metric for outcome differences. With two alternatives (A and B) under consideration, there are three possible conclusions: Option A is substantially better; option B is substantially better; and the options are more or less equal—that is, the decision is a close call or toss up (1). When considered to sufficient precision, it would be unusual for two strategies to be exactly equivalent. Thus, the boundary between a close call and a clear benefit or harm is dependent on the purpose of the comparison, the perspective of the analysis (who the decision maker is), and the preferences of the decision maker. In recent decades, such comparisons have been sharpened by developing models of the choice and the prognoses using decision analysis that explicitly reveals the underlying assumptions and perspectives. Prognosis and its uncertainty can be summarized by probabilities or probability distributions. The relative value of possible outcomes can be summarized by utilities or utility distributions. Sometimes utilities reflect not only hard data, such as survival, but also quality of life, which clearly varies based on whose preferences are being considered. Such decision models also require the explicit specification of the time horizon being considered (2). When the problem being represented involves evolution over time, it is commonplace to use a Markov or state transition model (3), which " follows " the prognosis of a cohort of similar patients over time as they move from one health state to another health state, with transitions governed by the probabilities derived from available data. Eventually the cohort gets absorbed in the so-called " Dead " state. The sum of the incremental utilities over the time frame of the model is used as a measure of the value of each strategy under consideration. Whether or not to undergo a contralateral prophylactic mas-tectomy (CPM) after being treated for breast cancer is a difficult choice for many women. The goal of such aggressive therapy is to lower the likelihood of a second primary carcinoma developing. The downsides are operative risk, impairment of the woman's self-image, and short-term and long-term morbidities. In this issue of the Journal, Portschy and colleagues from the University of Minnesota offer a Markov model of this choice (4). They consider average women with breast cancer and not …