Pitfalls in outcome prediction of breast cancer.

Breast cancer represents a heterogeneous group of diseases with varied presentation, morphological and biological features, behaviour, and response to therapy. Management of breast cancer relies on availability of robust predictive and prognostic factors to support decision making. Identifying and validating the prognostic and predictive value of a given marker is based on studying its association with clinical outcome with or without consideration for therapy, respectively. In the field of cancer research, clinical outcome is determined by assessing certain time-dependent events: 'endpoints' such as tumour progression, recurrence and patient mortality. Guidelines for reporting tumour markers have been published and there is a perception that outcome determination in breast cancer is well documented. However, reviewing the literature has highlighted the varied use of definitions used in clinical outcome measures and there are pitfalls in outcome analysis. This may have contributed to the discrepancies in the literature and to the inconsistent conclusions seen in published studies assessing the same markers. Identification of these pitfalls is expected to improve prognostic and predictive marker assessment. Here issues related to outcome determination in breast cancer including definitions and pitfalls and some critical views are presented.