Prognostic Models: A Methodological Framework and Review of Models for Breast Cancer

Prognostic models are widely used in cancer for investigating patient outcome in relation to multiple patient and disease characteristics. Such a model may allow the (reasonably) reliable classification of patients into two or more groups with different prognoses. It may be of particular interest to identify patients with a good prognosis that adjuvant therapy would not be (cost-)beneficial, or a group with a poor prognosis that more aggressive adjuvant therapy would not be justified (1). We can define a prognostic model as a combination of at least two separate variables to predict patient outcome. In this chapter, I describe a review of published prognostic models that have been developed to predict the outcome of future breast cancer patients. Studies were considered where the specific aim was either to develop a new prognostic model or to attempt to validate an existing model for newly diagnosed patients, and where at least one of the endpoints of death, cancer death, or recurrence of disease was studied. A full account is available elsewhere (2). It is clear that a prognostic model will have no clinical value unless it can be shown to predict outcome with some success; unless the model is shown to be useful it will be quickly forgotten (3). Thus, there is a particular interest in identifying prognostic models for which there has been an evaluation of how successful the models have been when used in a different setting, that is, models which have been validated externally (1, 4, 5).