Prognostic or Predictive? It's Time to Get Back to Definitions!

TO THEEDITOR: I read with great interest the study of Peck et al 1 assessing the correlation of nuclear localized and tyrosine phosphorylated Stat5 expression in breast cancer with outcome. The authors claimed in the Discussion section that their results “revealed a strong predictive value of Nuc-pYStat5 for response to antiestrogen therapy.” I believe that this statement is misleading. In the literature, there is frequently some confusion about the definitions of the terms “prognostic” and “predictive.” A prognostic factor is a clinical or biologic characteristic that is objectively measurable and that provides information on the likely outcome of the cancer disease in an untreated individual. Such prognostic markers are helpful for identifying patients with cancer who are at high risk of metastatic relapse and therefore potential candidates for adjuvant systemic treatments. In contrast, a predictive factor is a clinical or biologic characteristic that provides information on the likely benefit from treatment (either in terms of tumor shrinkage or survival). Such predictive factors can be used to identify subpopulations of patients who are most likely to benefit from a given therapy. Importantly, prognostic factors define the effects of patient or tumor characteristics on the patient outcome, whereas predictive factors define the effect of treatment on the tumor. By analyzing the impact of Stat5 status on the survival of patients with lymph node–negative breast cancer who did not receive adjuvant therapy, the authors have clearly demonstrated the prognostic value of this biomarker in this setting. Indeed, patients with low Nuc-pYStat5 expression had a significantly higher risk of disease recurrence and of dying from breast cancer than patients with high Nuc-pYStat5 expression. To analyze the potential predictive value of Stat5 status for response to estrogen therapy, the authors used the same methodology in patients with node-negative and node-positive breast cancer who received adjuvant antiestrogen therapy. They found a significant correlation between low levels of Nuc-pYStat5 and poor breast cancer– specific survival and therefore concluded that Stat5 expression status was predictive of response to hormonal therapy. This conclusion is not accurate. By using such a methodology, the authors proved only that Stat5 status also had a prognostic value in patients managed with antiestrogen. However, this does not indicate that Stat5 can be used as a predictive biomarker to select patients who are more likely to benefit from hormonal treatment. The only way to demonstrate a predictive value of Stat5 status for benefit of hormonal treatment in the adjuvant setting would be to retrospectively or prospectively analyze the data from a randomized controlled trial comparing hormonal therapy versus placebo or versus a regimen devoid of antiestrogen drugs. Only with that kind of study can it be demonstrated that the benefit of hormonal therapy is limited to patients with high Stat5 expression and that Stat5 status can even be considered as a predictive factor. Although prognostic marker validation is relatively easy and can be established by using data from retrospective series, more stringent criteria are required for the validation of predictive biomarkers. 2 With regard to the predictive value of Stat5 status for hormonal therapy in breast cancer, the study design chosen by Peck et al 1 does not fit these criteria.